commit 1f880799a49a8f0160f5871385dfebc1aae57c78 Author: ninghongbin <2409766686@qq.com> Date: Thu Oct 16 17:18:10 2025 +0800 盒子ocr检测 diff --git a/.gitignore b/.gitignore new file mode 100644 index 0000000..82f9275 --- /dev/null +++ b/.gitignore @@ -0,0 +1,162 @@ +# Byte-compiled / optimized / DLL files +__pycache__/ +*.py[cod] +*$py.class + +# C extensions +*.so + +# Distribution / packaging +.Python +build/ +develop-eggs/ +dist/ +downloads/ +eggs/ +.eggs/ +lib/ +lib64/ +parts/ +sdist/ +var/ +wheels/ +share/python-wheels/ +*.egg-info/ +.installed.cfg +*.egg +MANIFEST + +# PyInstaller +# Usually these files are written by a python script from a template +# before PyInstaller builds the exe, so as to inject date/other infos into it. +*.manifest +*.spec + +# Installer logs +pip-log.txt +pip-delete-this-directory.txt + +# Unit test / coverage reports +htmlcov/ +.tox/ +.nox/ +.coverage +.coverage.* +.cache +nosetests.xml +coverage.xml +*.cover +*.py,cover +.hypothesis/ +.pytest_cache/ +cover/ + +# Translations +*.mo +*.pot + +# Django stuff: +*.log +local_settings.py +db.sqlite3 +db.sqlite3-journal + +# Flask stuff: +instance/ +.webassets-cache + +# Scrapy stuff: +.scrapy + +# Sphinx documentation +docs/_build/ + +# PyBuilder +.pybuilder/ +target/ + +# Jupyter Notebook +.ipynb_checkpoints + +# IPython +profile_default/ +ipython_config.py + +# pyenv +# For a library or package, you might want to ignore these files since the code is +# intended to run in multiple environments; otherwise, check them in: +# .python-version + +# pipenv +# According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control. +# However, in case of collaboration, if having platform-specific dependencies or dependencies +# having no cross-platform support, pipenv may install dependencies that don't work, or not +# install all needed dependencies. +#Pipfile.lock + +# poetry +# Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control. +# This is especially recommended for binary packages to ensure reproducibility, and is more +# commonly ignored for libraries. +# https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control +#poetry.lock + +# pdm +# Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control. +#pdm.lock +# pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it +# in version control. +# https://pdm.fming.dev/latest/usage/project/#working-with-version-control +.pdm.toml +.pdm-python +.pdm-build/ + +# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm +__pypackages__/ + +# Celery stuff +celerybeat-schedule +celerybeat.pid + +# SageMath parsed files +*.sage.py + +# Environments +.env +.venv +env/ +venv/ +ENV/ +env.bak/ +venv.bak/ + +# Spyder project settings +.spyderproject +.spyproject + +# Rope project settings +.ropeproject + +# mkdocs documentation +/site + +# mypy +.mypy_cache/ +.dmypy.json +dmypy.json + +# Pyre type checker +.pyre/ + +# pytype static type analyzer +.pytype/ + +# Cython debug symbols +cython_debug/ + +# PyCharm +# JetBrains specific template is maintained in a separate JetBrains.gitignore that can +# be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore +# and can be added to the global gitignore or merged into this file. For a more nuclear +# option (not recommended) you can uncomment the following to ignore the entire idea folder. +#.idea/ diff --git a/.idea/.gitignore b/.idea/.gitignore new file mode 100644 index 0000000..35410ca --- /dev/null +++ b/.idea/.gitignore @@ -0,0 +1,8 @@ +# 默认忽略的文件 +/shelf/ +/workspace.xml +# 基于编辑器的 HTTP 客户端请求 +/httpRequests/ +# Datasource local storage ignored files +/dataSources/ +/dataSources.local.xml diff --git a/.idea/inspectionProfiles/profiles_settings.xml b/.idea/inspectionProfiles/profiles_settings.xml new file mode 100644 index 0000000..105ce2d --- /dev/null +++ b/.idea/inspectionProfiles/profiles_settings.xml @@ -0,0 +1,6 @@ + + + + \ No newline at end of file diff --git a/.idea/misc.xml b/.idea/misc.xml new file mode 100644 index 0000000..3dc10fe --- /dev/null +++ b/.idea/misc.xml @@ -0,0 +1,7 @@ + + + + + + \ No newline at end of file diff --git a/.idea/modules.xml b/.idea/modules.xml new file mode 100644 index 0000000..a2d0614 --- /dev/null +++ b/.idea/modules.xml @@ -0,0 +1,8 @@ + + + + + + + + \ No newline at end of file diff --git a/.idea/pp_onnx-main.iml b/.idea/pp_onnx-main.iml new file mode 100644 index 0000000..643bee7 --- /dev/null +++ b/.idea/pp_onnx-main.iml @@ -0,0 +1,12 @@ + + + + + + + + + + \ No newline at end of file diff --git a/1 b/1 new file mode 100644 index 0000000..55b1829 --- /dev/null +++ b/1 @@ -0,0 +1,255 @@ +文件一：test_ocr +import cv2 +import time +from pp_onnx.onnx_paddleocr import ONNXPaddleOcr, draw_ocr + +model = ONNXPaddleOcr( + use_angle_cls=True, + use_gpu=False, + providers=['RKNNExecutionProvider'], + provider_options=[{'device_id': 0}] +) + +try: + # 获取文本检测模型的ONNX会话 + onnx_session = model.det_session + # 获取实际使用的执行提供者 + used_providers = onnx_session.get_providers() + print(f"当前使用的执行提供者（计算设备）：{used_providers}") + + if 'RKNNExecutionProvider' in used_providers: + print("✅ 成功使用RK3588 NPU加速推理") + else: + print("❌ 未使用NPU，当前设备：", used_providers) +except AttributeError as e: + print(f"获取会话失败：{e}，请检查 onnx_paddleocr.py 中会话属性名是否正确（如 det_session/rec_session）") + + +def sav2Img(org_img, result, name="./result_img/draw_ocr_996_1.jpg"): + from PIL import Image + result = result[0] + image = org_img[:, :, ::-1] + boxes = [line[0] for line in result] + txts = [line[1][0] for line in result] + scores = [line[1][1] for line in result] + im_show = draw_ocr(image, boxes, txts, scores) + im_show = Image.fromarray(im_show) + im_show.save(name) + + +# 执行OCR推理 +img = cv2.imread('./test_img/test1.jpg') +if img is None: + print(f"❌ 未找到图像文件：./test_img/test1.jpg") +else: + s = time.time() + result = model.ocr(img) + e = time.time() + print(f"total time: {e - s:.3f} 秒") + print("result:", result) + for box in result[0]: + print(box) + sav2Img(img, result) + + +文件二：onnx_paddleocr +import time + +from pp_onnx.predict_system import TextSystem +from pp_onnx.utils import infer_args as init_args +from pp_onnx.utils import str2bool, draw_ocr +import argparse +import sys + + +class ONNXPaddleOcr(TextSystem): + def __init__(self, **kwargs): + # 默认参数 + parser = init_args() + # import IPython + # IPython.embed(header='L-14') + + inference_args_dict = {} + for action in parser._actions: + inference_args_dict[action.dest] = action.default + params = argparse.Namespace(**inference_args_dict) + + params.rec_image_shape = "3, 48, 320" + + # 根据传入的参数覆盖更新默认参数 + params.__dict__.update(**kwargs) + + # 初始化模型 + super().__init__(params) + + def ocr(self, img, det=True, rec=True, cls=True): + if cls == True and self.use_angle_cls == False: + print('Since the angle classifier is not initialized, the angle classifier will not be uesd during the forward process') + + if det and rec: + ocr_res = [] + dt_boxes, rec_res = self.__call__(img, cls) + tmp_res = [[box.tolist(), res] for box, res in zip(dt_boxes, rec_res)] + ocr_res.append(tmp_res) + return ocr_res + elif det and not rec: + ocr_res = [] + dt_boxes = self.text_detector(img) + tmp_res = [box.tolist() for box in dt_boxes] + ocr_res.append(tmp_res) + return ocr_res + else: + ocr_res = [] + cls_res = [] + + if not isinstance(img, list): + img = [img] + if self.use_angle_cls and cls: + img, cls_res_tmp = self.text_classifier(img) + if not rec: + cls_res.append(cls_res_tmp) + rec_res = self.text_recognizer(img) + ocr_res.append(rec_res) + + if not rec: + return cls_res + return ocr_res + + +def sav2Img(org_img, result, name="draw_ocr.jpg"): + # 显示结果 + from PIL import Image + result = result[0] + # image = Image.open(img_path).convert('RGB') + # 图像转BGR2RGB + image = org_img[:, :, ::-1] + boxes = [line[0] for line in result] + txts = [line[1][0] for line in result] + scores = [line[1][1] for line in result] + im_show = draw_ocr(image, boxes, txts, scores) + im_show = Image.fromarray(im_show) + im_show.save(name) + + +if __name__ == '__main__': + import cv2 + + model = ONNXPaddleOcr(use_angle_cls=True, use_gpu=False) + + + img = cv2.imread('/data2/liujingsong3/fiber_box/test/img/20230531230052008263304.jpg') + s = time.time() + result = model.ocr(img) + e = time.time() + print("total time: {:.3f}".format(e - s)) + print("result:", result) + for box in result[0]: + print(box) + + sav2Img(img, result) + + +文件三：predict_system + +import os +import cv2 +import copy +import pp_onnx.predict_det as predict_det +import pp_onnx.predict_cls as predict_cls +import pp_onnx.predict_rec as predict_rec +from pp_onnx.utils import get_rotate_crop_image, get_minarea_rect_crop + + +class TextSystem(object): + def __init__(self, args): + self.text_detector = predict_det.TextDetector(args) + self.text_recognizer = predict_rec.TextRecognizer(args) + self.use_angle_cls = args.use_angle_cls + self.drop_score = args.drop_score + if self.use_angle_cls: + self.text_classifier = predict_cls.TextClassifier(args) + + self.args = args + self.crop_image_res_index = 0 + + + def draw_crop_rec_res(self, output_dir, img_crop_list, rec_res): + os.makedirs(output_dir, exist_ok=True) + bbox_num = len(img_crop_list) + for bno in range(bbox_num): + cv2.imwrite( + os.path.join(output_dir, + f"mg_crop_{bno+self.crop_image_res_index}.jpg"), + img_crop_list[bno]) + + self.crop_image_res_index += bbox_num + + def __call__(self, img, cls=True): + ori_im = img.copy() + # 文字检测 + dt_boxes = self.text_detector(img) + + if dt_boxes is None: + return None, None + + img_crop_list = [] + + dt_boxes = sorted_boxes(dt_boxes) + + # 图片裁剪 + for bno in range(len(dt_boxes)): + tmp_box = copy.deepcopy(dt_boxes[bno]) + if self.args.det_box_type == "quad": + img_crop = get_rotate_crop_image(ori_im, tmp_box) + else: + img_crop = get_minarea_rect_crop(ori_im, tmp_box) + img_crop_list.append(img_crop) + + # 方向分类 + if self.use_angle_cls and cls: + img_crop_list, angle_list = self.text_classifier(img_crop_list) + + # 图像识别 + rec_res = self.text_recognizer(img_crop_list) + + if self.args.save_crop_res: + self.draw_crop_rec_res(self.args.crop_res_save_dir, img_crop_list,rec_res) + filter_boxes, filter_rec_res = [], [] + for box, rec_result in zip(dt_boxes, rec_res): + text, score = rec_result + if score >= self.drop_score: + filter_boxes.append(box) + filter_rec_res.append(rec_result) + + # import IPython + # IPython.embed(header='L-70') + + return filter_boxes, filter_rec_res + + +def sorted_boxes(dt_boxes): + """ + Sort text boxes in order from top to bottom, left to right + args: + dt_boxes(array):detected text boxes with shape [4, 2] + return: + sorted boxes(array) with shape [4, 2] + """ + num_boxes = dt_boxes.shape[0] + sorted_boxes = sorted(dt_boxes, key=lambda x: (x[0][1], x[0][0])) + _boxes = list(sorted_boxes) + + for i in range(num_boxes - 1): + for j in range(i, -1, -1): + if abs(_boxes[j + 1][0][1] - _boxes[j][0][1]) < 10 and \ + (_boxes[j + 1][0][0] < _boxes[j][0][0]): + tmp = _boxes[j] + _boxes[j] = _boxes[j + 1] + _boxes[j + 1] = tmp + else: + break + return _boxes + +运行test_ocr报错：(box_ocr) root@ztl:/result/ocr/pp_onnx-main# python test_ocr.py +获取会话失败：'ONNXPaddleOcr' object has no attribute 'det_session'，请检查 onnx_paddleocr.py 中会话属性名是否正确（如 det_session/rec_session） +total time: 11.161 秒 且检测一张图片耗时太差 \ No newline at end of file diff --git a/1.jpg b/1.jpg new file mode 100644 index 0000000..1f15a3c Binary files /dev/null and b/1.jpg differ diff --git a/LICENSE b/LICENSE new file mode 100644 index 0000000..261eeb9 --- /dev/null +++ b/LICENSE @@ -0,0 +1,201 @@ + Apache License + Version 2.0, January 2004 + http://www.apache.org/licenses/ + + TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION + + 1. Definitions. + + "License" shall mean the terms and conditions for use, reproduction, + and distribution as defined by Sections 1 through 9 of this document. + + "Licensor" shall mean the copyright owner or entity authorized by + the copyright owner that is granting the License. + + "Legal Entity" shall mean the union of the acting entity and all + other entities that control, are controlled by, or are under common + control with that entity. For the purposes of this definition, + "control" means (i) the power, direct or indirect, to cause the + direction or management of such entity, whether by contract or + otherwise, or (ii) ownership of fifty percent (50%) or more of the + outstanding shares, or (iii) beneficial ownership of such entity. + + "You" (or "Your") shall mean an individual or Legal Entity + exercising permissions granted by this License. + + "Source" form shall mean the preferred form for making modifications, + including but not limited to software source code, documentation + source, and configuration files. + + "Object" form shall mean any form resulting from mechanical + transformation or translation of a Source form, including but + not limited to compiled object code, generated documentation, + and conversions to other media types. + + "Work" shall mean the work of authorship, whether in Source or + Object form, made available under the License, as indicated by a + copyright notice that is included in or attached to the work + (an example is provided in the Appendix below). + + "Derivative Works" shall mean any work, whether in Source or Object + form, that is based on (or derived from) the Work and for which the + editorial revisions, annotations, elaborations, or other modifications + represent, as a whole, an original work of authorship. For the purposes + of this License, Derivative Works shall not include works that remain + separable from, or merely link (or bind by name) to the interfaces of, + the Work and Derivative Works thereof. + + "Contribution" shall mean any work of authorship, including + the original version of the Work and any modifications or additions + to that Work or Derivative Works thereof, that is intentionally + submitted to Licensor for inclusion in the Work by the copyright owner + or by an individual or Legal Entity authorized to submit on behalf of + the copyright owner. For the purposes of this definition, "submitted" + means any form of electronic, verbal, or written communication sent + to the Licensor or its representatives, including but not limited to + communication on electronic mailing lists, source code control systems, + and issue tracking systems that are managed by, or on behalf of, the + Licensor for the purpose of discussing and improving the Work, but + excluding communication that is conspicuously marked or otherwise + designated in writing by the copyright owner as "Not a Contribution." + + "Contributor" shall mean Licensor and any individual or Legal Entity + on behalf of whom a Contribution has been received by Licensor and + subsequently incorporated within the Work. + + 2. Grant of Copyright License. Subject to the terms and conditions of + this License, each Contributor hereby grants to You a perpetual, + worldwide, non-exclusive, no-charge, royalty-free, irrevocable + copyright license to reproduce, prepare Derivative Works of, + publicly display, publicly perform, sublicense, and distribute the + Work and such Derivative Works in Source or Object form. + + 3. Grant of Patent License. Subject to the terms and conditions of + this License, each Contributor hereby grants to You a perpetual, + worldwide, non-exclusive, no-charge, royalty-free, irrevocable + (except as stated in this section) patent license to make, have made, + use, offer to sell, sell, import, and otherwise transfer the Work, + where such license applies only to those patent claims licensable + by such Contributor that are necessarily infringed by their + Contribution(s) alone or by combination of their Contribution(s) + with the Work to which such Contribution(s) was submitted. If You + institute patent litigation against any entity (including a + cross-claim or counterclaim in a lawsuit) alleging that the Work + or a Contribution incorporated within the Work constitutes direct + or contributory patent infringement, then any patent licenses + granted to You under this License for that Work shall terminate + as of the date such litigation is filed. + + 4. Redistribution. You may reproduce and distribute copies of the + Work or Derivative Works thereof in any medium, with or without + modifications, and in Source or Object form, provided that You + meet the following conditions: + + (a) You must give any other recipients of the Work or + Derivative Works a copy of this License; and + + (b) You must cause any modified files to carry prominent notices + stating that You changed the files; and + + (c) You must retain, in the Source form of any Derivative Works + that You distribute, all copyright, patent, trademark, and + attribution notices from the Source form of the Work, + excluding those notices that do not pertain to any part of + the Derivative Works; and + + (d) If the Work includes a "NOTICE" text file as part of its + distribution, then any Derivative Works that You distribute must + include a readable copy of the attribution notices contained + within such NOTICE file, excluding those notices that do not + pertain to any part of the Derivative Works, in at least one + of the following places: within a NOTICE text file distributed + as part of the Derivative Works; within the Source form or + documentation, if provided along with the Derivative Works; or, + within a display generated by the Derivative Works, if and + wherever such third-party notices normally appear. The contents + of the NOTICE file are for informational purposes only and + do not modify the License. You may add Your own attribution + notices within Derivative Works that You distribute, alongside + or as an addendum to the NOTICE text from the Work, provided + that such additional attribution notices cannot be construed + as modifying the License. + + You may add Your own copyright statement to Your modifications and + may provide additional or different license terms and conditions + for use, reproduction, or distribution of Your modifications, or + for any such Derivative Works as a whole, provided Your use, + reproduction, and distribution of the Work otherwise complies with + the conditions stated in this License. + + 5. Submission of Contributions. Unless You explicitly state otherwise, + any Contribution intentionally submitted for inclusion in the Work + by You to the Licensor shall be under the terms and conditions of + this License, without any additional terms or conditions. + Notwithstanding the above, nothing herein shall supersede or modify + the terms of any separate license agreement you may have executed + with Licensor regarding such Contributions. + + 6. Trademarks. This License does not grant permission to use the trade + names, trademarks, service marks, or product names of the Licensor, + except as required for reasonable and customary use in describing the + origin of the Work and reproducing the content of the NOTICE file. + + 7. Disclaimer of Warranty. Unless required by applicable law or + agreed to in writing, Licensor provides the Work (and each + Contributor provides its Contributions) on an "AS IS" BASIS, + WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or + implied, including, without limitation, any warranties or conditions + of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A + PARTICULAR PURPOSE. You are solely responsible for determining the + appropriateness of using or redistributing the Work and assume any + risks associated with Your exercise of permissions under this License. + + 8. Limitation of Liability. In no event and under no legal theory, + whether in tort (including negligence), contract, or otherwise, + unless required by applicable law (such as deliberate and grossly + negligent acts) or agreed to in writing, shall any Contributor be + liable to You for damages, including any direct, indirect, special, + incidental, or consequential damages of any character arising as a + result of this License or out of the use or inability to use the + Work (including but not limited to damages for loss of goodwill, + work stoppage, computer failure or malfunction, or any and all + other commercial damages or losses), even if such Contributor + has been advised of the possibility of such damages. + + 9. Accepting Warranty or Additional Liability. While redistributing + the Work or Derivative Works thereof, You may choose to offer, + and charge a fee for, acceptance of support, warranty, indemnity, + or other liability obligations and/or rights consistent with this + License. However, in accepting such obligations, You may act only + on Your own behalf and on Your sole responsibility, not on behalf + of any other Contributor, and only if You agree to indemnify, + defend, and hold each Contributor harmless for any liability + incurred by, or claims asserted against, such Contributor by reason + of your accepting any such warranty or additional liability. + + END OF TERMS AND CONDITIONS + + APPENDIX: How to apply the Apache License to your work. + + To apply the Apache License to your work, attach the following + boilerplate notice, with the fields enclosed by brackets "[]" + replaced with your own identifying information. (Don't include + the brackets!) The text should be enclosed in the appropriate + comment syntax for the file format. We also recommend that a + file or class name and description of purpose be included on the + same "printed page" as the copyright notice for easier + identification within third-party archives. + + Copyright [yyyy] [name of copyright owner] + + Licensed under the Apache License, Version 2.0 (the "License"); + you may not use this file except in compliance with the License. + You may obtain a copy of the License at + + http://www.apache.org/licenses/LICENSE-2.0 + + Unless required by applicable law or agreed to in writing, software + distributed under the License is distributed on an "AS IS" BASIS, + WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + See the License for the specific language governing permissions and + limitations under the License. diff --git a/Readme.md b/Readme.md new file mode 100644 index 0000000..d893def --- /dev/null +++ b/Readme.md @@ -0,0 +1,49 @@ +# onnxOCR +#### 一.优势： +1.脱离深度学习训练框架，可直接用于部署的通用OCR。 +2.在算力有限，精度不变的情况下使用paddleOCR转成ONNX模型，进行重新构建的一款可部署在arm架构和x86架构计算机上的OCR模型。 +3.在同样性能的计算机上推理速度加速了4-5倍。 + +#### 二.环境安装 + python>=3.6 + + pip install -i https://pypi.tuna.tsinghua.edu.cn/simple -r requirements.txt + + 由于rec模型超过了100M，github有限制，所以我上传到 +[百度网盘,提取码: 125c](https://pan.baidu.com/s/1O1b30CMwsDjD7Ti9EnxYKQ ) + + 下载后放到./models/ch_ppocr_server_v2.0/rec/rec.onnx下 + +#### 三.一键运行 + + python test_ocr.py + +#### 效果展示 + +![Alt text](result_img/draw_ocr_1.jpg) + +![Alt text](result_img/draw_ocr2.jpg) + +![Alt text](result_img/draw_ocr3.jpg) + +![Alt text](result_img/draw_ocr4.jpg) + +![Alt text](result_img/draw_ocr5.jpg) + +![Alt text](result_img/draw_ocr.jpg) + +#### 感谢PaddleOcr + +https://github.com/PaddlePaddle/PaddleOCR + +#### 从该项目Fork而来 +https://github.com/jingsongliujing/OnnxOCR + +--- + +CHANGELOG + +1. 加入最新的`pp_ocr_v4`的检测与识别模型 +2. 修改包名为`pp_onnx`防止与onnx冲突 +3. 修改部分写死的参数 + diff --git a/__init__.py b/__init__.py new file mode 100644 index 0000000..929a8c4 --- /dev/null +++ b/__init__.py @@ -0,0 +1,6 @@ +import os +import sys + +sys.path.append(os.path.dirname(os.path.abspath(__file__))) + +from pp_onnx.onnx_paddleocr import ONNXPaddleOcr \ No newline at end of file diff --git a/det_result.jpg b/det_result.jpg new file mode 100644 index 0000000..9d0a429 Binary files /dev/null and b/det_result.jpg differ diff --git a/pp_onnx/__init__.py b/pp_onnx/__init__.py new file mode 100644 index 0000000..e69de29 diff --git a/pp_onnx/cls_postprocess.py b/pp_onnx/cls_postprocess.py new file mode 100644 index 0000000..473758d --- /dev/null +++ b/pp_onnx/cls_postprocess.py @@ -0,0 +1,30 @@ + +# import paddle + + +class ClsPostProcess(object): + """ Convert between text-label and text-index """ + + def __init__(self, label_list=None, key=None, **kwargs): + super(ClsPostProcess, self).__init__() + self.label_list = label_list + self.key = key + + def __call__(self, preds, label=None, *args, **kwargs): + if self.key is not None: + preds = preds[self.key] + + label_list = self.label_list + if label_list is None: + label_list = {idx: idx for idx in range(preds.shape[-1])} + + # if isinstance(preds, paddle.Tensor): + # preds = preds.numpy() + + pred_idxs = preds.argmax(axis=1) + decode_out = [(label_list[idx], preds[i, idx]) + for i, idx in enumerate(pred_idxs)] + if label is None: + return decode_out + label = [(label_list[idx], 1.0) for idx in label] + return decode_out, label diff --git a/pp_onnx/db_postprocess.py b/pp_onnx/db_postprocess.py new file mode 100644 index 0000000..4f05fd9 --- /dev/null +++ b/pp_onnx/db_postprocess.py @@ -0,0 +1,276 @@ +# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +""" +This code is refered from: +https://github.com/WenmuZhou/DBNet.pytorch/blob/master/post_processing/seg_detector_representer.py +""" +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +import numpy as np +import cv2 +# import paddle +from shapely.geometry import Polygon +import pyclipper + + +class DBPostProcess(object): + """ + The post process for Differentiable Binarization (DB). + """ + + def __init__(self, + thresh=0.3, + box_thresh=0.7, + max_candidates=1000, + unclip_ratio=2.0, + use_dilation=False, + score_mode="fast", + box_type='quad', + **kwargs): + self.thresh = thresh + self.box_thresh = box_thresh + self.max_candidates = max_candidates + self.unclip_ratio = unclip_ratio + self.min_size = 3 + self.score_mode = score_mode + self.box_type = box_type + assert score_mode in [ + "slow", "fast" + ], "Score mode must be in [slow, fast] but got: {}".format(score_mode) + + self.dilation_kernel = None if not use_dilation else np.array( + [[1, 1], [1, 1]]) + + def polygons_from_bitmap(self, pred, _bitmap, dest_width, dest_height): + ''' + _bitmap: single map with shape (1, H, W), + whose values are binarized as {0, 1} + ''' + + bitmap = _bitmap + height, width = bitmap.shape + + boxes = [] + scores = [] + + contours, _ = cv2.findContours((bitmap * 255).astype(np.uint8), + cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE) + + for contour in contours[:self.max_candidates]: + epsilon = 0.002 * cv2.arcLength(contour, True) + approx = cv2.approxPolyDP(contour, epsilon, True) + points = approx.reshape((-1, 2)) + if points.shape[0] < 4: + continue + + score = self.box_score_fast(pred, points.reshape(-1, 2)) + if self.box_thresh > score: + continue + + if points.shape[0] > 2: + box = self.unclip(points, self.unclip_ratio) + if len(box) > 1: + continue + else: + continue + box = box.reshape(-1, 2) + + _, sside = self.get_mini_boxes(box.reshape((-1, 1, 2))) + if sside < self.min_size + 2: + continue + + box = np.array(box) + box[:, 0] = np.clip( + np.round(box[:, 0] / width * dest_width), 0, dest_width) + box[:, 1] = np.clip( + np.round(box[:, 1] / height * dest_height), 0, dest_height) + boxes.append(box.tolist()) + scores.append(score) + return boxes, scores + + def boxes_from_bitmap(self, pred, _bitmap, dest_width, dest_height): + ''' + _bitmap: single map with shape (1, H, W), + whose values are binarized as {0, 1} + ''' + + bitmap = _bitmap + height, width = bitmap.shape + + outs = cv2.findContours((bitmap * 255).astype(np.uint8), cv2.RETR_LIST, + cv2.CHAIN_APPROX_SIMPLE) + if len(outs) == 3: + img, contours, _ = outs[0], outs[1], outs[2] + elif len(outs) == 2: + contours, _ = outs[0], outs[1] + + num_contours = min(len(contours), self.max_candidates) + + boxes = [] + scores = [] + for index in range(num_contours): + contour = contours[index] + points, sside = self.get_mini_boxes(contour) + if sside < self.min_size: + continue + points = np.array(points) + if self.score_mode == "fast": + score = self.box_score_fast(pred, points.reshape(-1, 2)) + else: + score = self.box_score_slow(pred, contour) + if self.box_thresh > score: + continue + + box = self.unclip(points, self.unclip_ratio).reshape(-1, 1, 2) + box, sside = self.get_mini_boxes(box) + if sside < self.min_size + 2: + continue + box = np.array(box) + + box[:, 0] = np.clip( + np.round(box[:, 0] / width * dest_width), 0, dest_width) + box[:, 1] = np.clip( + np.round(box[:, 1] / height * dest_height), 0, dest_height) + boxes.append(box.astype("int32")) + scores.append(score) + return np.array(boxes, dtype="int32"), scores + + def unclip(self, box, unclip_ratio): + poly = Polygon(box) + distance = poly.area * unclip_ratio / poly.length + offset = pyclipper.PyclipperOffset() + offset.AddPath(box, pyclipper.JT_ROUND, pyclipper.ET_CLOSEDPOLYGON) + expanded = np.array(offset.Execute(distance)) + return expanded + + def get_mini_boxes(self, contour): + bounding_box = cv2.minAreaRect(contour) + points = sorted(list(cv2.boxPoints(bounding_box)), key=lambda x: x[0]) + + index_1, index_2, index_3, index_4 = 0, 1, 2, 3 + if points[1][1] > points[0][1]: + index_1 = 0 + index_4 = 1 + else: + index_1 = 1 + index_4 = 0 + if points[3][1] > points[2][1]: + index_2 = 2 + index_3 = 3 + else: + index_2 = 3 + index_3 = 2 + + box = [ + points[index_1], points[index_2], points[index_3], points[index_4] + ] + return box, min(bounding_box[1]) + + def box_score_fast(self, bitmap, _box): + ''' + box_score_fast: use bbox mean score as the mean score + ''' + h, w = bitmap.shape[:2] + box = _box.copy() + xmin = np.clip(np.floor(box[:, 0].min()).astype("int32"), 0, w - 1) + xmax = np.clip(np.ceil(box[:, 0].max()).astype("int32"), 0, w - 1) + ymin = np.clip(np.floor(box[:, 1].min()).astype("int32"), 0, h - 1) + ymax = np.clip(np.ceil(box[:, 1].max()).astype("int32"), 0, h - 1) + + mask = np.zeros((ymax - ymin + 1, xmax - xmin + 1), dtype=np.uint8) + box[:, 0] = box[:, 0] - xmin + box[:, 1] = box[:, 1] - ymin + cv2.fillPoly(mask, box.reshape(1, -1, 2).astype("int32"), 1) + return cv2.mean(bitmap[ymin:ymax + 1, xmin:xmax + 1], mask)[0] + + def box_score_slow(self, bitmap, contour): + ''' + box_score_slow: use polyon mean score as the mean score + ''' + h, w = bitmap.shape[:2] + contour = contour.copy() + contour = np.reshape(contour, (-1, 2)) + + xmin = np.clip(np.min(contour[:, 0]), 0, w - 1) + xmax = np.clip(np.max(contour[:, 0]), 0, w - 1) + ymin = np.clip(np.min(contour[:, 1]), 0, h - 1) + ymax = np.clip(np.max(contour[:, 1]), 0, h - 1) + + mask = np.zeros((ymax - ymin + 1, xmax - xmin + 1), dtype=np.uint8) + + contour[:, 0] = contour[:, 0] - xmin + contour[:, 1] = contour[:, 1] - ymin + + cv2.fillPoly(mask, contour.reshape(1, -1, 2).astype("int32"), 1) + return cv2.mean(bitmap[ymin:ymax + 1, xmin:xmax + 1], mask)[0] + + def __call__(self, outs_dict, shape_list): + pred = outs_dict['maps'] + # if isinstance(pred, paddle.Tensor): + # pred = pred.numpy() + pred = pred[:, 0, :, :] + segmentation = pred > self.thresh + + boxes_batch = [] + for batch_index in range(pred.shape[0]): + src_h, src_w, ratio_h, ratio_w = shape_list[batch_index] + if self.dilation_kernel is not None: + mask = cv2.dilate( + np.array(segmentation[batch_index]).astype(np.uint8), + self.dilation_kernel) + else: + mask = segmentation[batch_index] + if self.box_type == 'poly': + boxes, scores = self.polygons_from_bitmap(pred[batch_index], + mask, src_w, src_h) + elif self.box_type == 'quad': + boxes, scores = self.boxes_from_bitmap(pred[batch_index], mask, + src_w, src_h) + else: + raise ValueError("box_type can only be one of ['quad', 'poly']") + + boxes_batch.append({'points': boxes}) + return boxes_batch + + +class DistillationDBPostProcess(object): + def __init__(self, + model_name=["student"], + key=None, + thresh=0.3, + box_thresh=0.6, + max_candidates=1000, + unclip_ratio=1.5, + use_dilation=False, + score_mode="fast", + box_type='quad', + **kwargs): + self.model_name = model_name + self.key = key + self.post_process = DBPostProcess( + thresh=thresh, + box_thresh=box_thresh, + max_candidates=max_candidates, + unclip_ratio=unclip_ratio, + use_dilation=use_dilation, + score_mode=score_mode, + box_type=box_type) + + def __call__(self, predicts, shape_list): + results = {} + for k in self.model_name: + results[k] = self.post_process(predicts[k], shape_list=shape_list) + return results diff --git a/pp_onnx/fonts/simfang.ttf b/pp_onnx/fonts/simfang.ttf new file mode 100644 index 0000000..2b59eae Binary files /dev/null and b/pp_onnx/fonts/simfang.ttf differ diff --git a/pp_onnx/imaug.py b/pp_onnx/imaug.py new file mode 100644 index 0000000..6a2dd77 --- /dev/null +++ b/pp_onnx/imaug.py @@ -0,0 +1,32 @@ +from pp_onnx.operators import * + +def transform(data, ops=None): + """ transform """ + if ops is None: + ops = [] + for op in ops: + data = op(data) + if data is None: + return None + return data + + +def create_operators(op_param_list, global_config=None): + """ + create operators based on the config + + Args: + params(list): a dict list, used to create some operators + """ + assert isinstance(op_param_list, list), ('operator config should be a list') + ops = [] + for operator in op_param_list: + assert isinstance(operator, + dict) and len(operator) == 1, "yaml format error" + op_name = list(operator)[0] + param = {} if operator[op_name] is None else operator[op_name] + if global_config is not None: + param.update(global_config) + op = eval(op_name)(**param) + ops.append(op) + return ops \ No newline at end of file diff --git a/pp_onnx/legancy/utils copy.py b/pp_onnx/legancy/utils copy.py new file mode 100644 index 0000000..a8b7320 --- /dev/null +++ b/pp_onnx/legancy/utils copy.py @@ -0,0 +1,341 @@ +import numpy as np +import cv2 +import argparse +import math +from PIL import Image, ImageDraw, ImageFont + +def get_rotate_crop_image(img, points): + ''' + img_height, img_width = img.shape[0:2] + left = int(np.min(points[:, 0])) + right = int(np.max(points[:, 0])) + top = int(np.min(points[:, 1])) + bottom = int(np.max(points[:, 1])) + img_crop = img[top:bottom, left:right, :].copy() + points[:, 0] = points[:, 0] - left + points[:, 1] = points[:, 1] - top + ''' + assert len(points) == 4, "shape of points must be 4*2" + img_crop_width = int( + max( + np.linalg.norm(points[0] - points[1]), + np.linalg.norm(points[2] - points[3]))) + img_crop_height = int( + max( + np.linalg.norm(points[0] - points[3]), + np.linalg.norm(points[1] - points[2]))) + pts_std = np.float32([[0, 0], [img_crop_width, 0], + [img_crop_width, img_crop_height], + [0, img_crop_height]]) + M = cv2.getPerspectiveTransform(points, pts_std) + dst_img = cv2.warpPerspective( + img, + M, (img_crop_width, img_crop_height), + borderMode=cv2.BORDER_REPLICATE, + flags=cv2.INTER_CUBIC) + dst_img_height, dst_img_width = dst_img.shape[0:2] + if dst_img_height * 1.0 / dst_img_width >= 1.5: + dst_img = np.rot90(dst_img) + return dst_img + +def get_minarea_rect_crop(img, points): + bounding_box = cv2.minAreaRect(np.array(points).astype(np.int32)) + points = sorted(list(cv2.boxPoints(bounding_box)), key=lambda x: x[0]) + + index_a, index_b, index_c, index_d = 0, 1, 2, 3 + if points[1][1] > points[0][1]: + index_a = 0 + index_d = 1 + else: + index_a = 1 + index_d = 0 + if points[3][1] > points[2][1]: + index_b = 2 + index_c = 3 + else: + index_b = 3 + index_c = 2 + + box = [points[index_a], points[index_b], points[index_c], points[index_d]] + crop_img = get_rotate_crop_image(img, np.array(box)) + return crop_img + + +def resize_img(img, input_size=600): + """ + resize img and limit the longest side of the image to input_size + """ + img = np.array(img) + im_shape = img.shape + im_size_max = np.max(im_shape[0:2]) + im_scale = float(input_size) / float(im_size_max) + img = cv2.resize(img, None, None, fx=im_scale, fy=im_scale) + return img + +def str_count(s): + """ + Count the number of Chinese characters, + a single English character and a single number + equal to half the length of Chinese characters. + args: + s(string): the input of string + return(int): + the number of Chinese characters + """ + import string + count_zh = count_pu = 0 + s_len = len(str(s)) + en_dg_count = 0 + for c in str(s): + if c in string.ascii_letters or c.isdigit() or c.isspace(): + en_dg_count += 1 + elif c.isalpha(): + count_zh += 1 + else: + count_pu += 1 + return s_len - math.ceil(en_dg_count / 2) + +def text_visual(texts, + scores, + img_h=400, + img_w=600, + threshold=0., + font_path="./fonts/simfang.ttf"): + """ + create new blank img and draw txt on it + args: + texts(list): the text will be draw + scores(list|None): corresponding score of each txt + img_h(int): the height of blank img + img_w(int): the width of blank img + font_path: the path of font which is used to draw text + return(array): + """ + if scores is not None: + assert len(texts) == len( + scores), "The number of txts and corresponding scores must match" + + def create_blank_img(): + blank_img = np.ones(shape=[img_h, img_w], dtype=np.int8) * 255 + blank_img[:, img_w - 1:] = 0 + blank_img = Image.fromarray(blank_img).convert("RGB") + draw_txt = ImageDraw.Draw(blank_img) + return blank_img, draw_txt + + blank_img, draw_txt = create_blank_img() + + font_size = 20 + txt_color = (0, 0, 0) + # import IPython; IPython.embed(header='L-129') + font = ImageFont.truetype(font_path, font_size, encoding="utf-8") + + gap = font_size + 5 + txt_img_list = [] + count, index = 1, 0 + for idx, txt in enumerate(texts): + index += 1 + if scores[idx] < threshold or math.isnan(scores[idx]): + index -= 1 + continue + first_line = True + while str_count(txt) >= img_w // font_size - 4: + tmp = txt + txt = tmp[:img_w // font_size - 4] + if first_line: + new_txt = str(index) + ': ' + txt + first_line = False + else: + new_txt = ' ' + txt + draw_txt.text((0, gap * count), new_txt, txt_color, font=font) + txt = tmp[img_w // font_size - 4:] + if count >= img_h // gap - 1: + txt_img_list.append(np.array(blank_img)) + blank_img, draw_txt = create_blank_img() + count = 0 + count += 1 + if first_line: + new_txt = str(index) + ': ' + txt + ' ' + '%.3f' % (scores[idx]) + else: + new_txt = " " + txt + " " + '%.3f' % (scores[idx]) + draw_txt.text((0, gap * count), new_txt, txt_color, font=font) + # whether add new blank img or not + if count >= img_h // gap - 1 and idx + 1 < len(texts): + txt_img_list.append(np.array(blank_img)) + blank_img, draw_txt = create_blank_img() + count = 0 + count += 1 + txt_img_list.append(np.array(blank_img)) + if len(txt_img_list) == 1: + blank_img = np.array(txt_img_list[0]) + else: + blank_img = np.concatenate(txt_img_list, axis=1) + return np.array(blank_img) + +def draw_ocr(image, + boxes, + txts=None, + scores=None, + drop_score=0.5, + font_path="./pp_onnx/fonts/simfang.ttf"): + """ + Visualize the results of OCR detection and recognition + args: + image(Image|array): RGB image + boxes(list): boxes with shape(N, 4, 2) + txts(list): the texts + scores(list): txxs corresponding scores + drop_score(float): only scores greater than drop_threshold will be visualized + font_path: the path of font which is used to draw text + return(array): + the visualized img + """ + if scores is None: + scores = [1] * len(boxes) + box_num = len(boxes) + for i in range(box_num): + if scores is not None and (scores[i] < drop_score or + math.isnan(scores[i])): + continue + box = np.reshape(np.array(boxes[i]), [-1, 1, 2]).astype(np.int64) + image = cv2.polylines(np.array(image), [box], True, (255, 0, 0), 2) + if txts is not None: + img = np.array(resize_img(image, input_size=600)) + txt_img = text_visual( + txts, + scores, + img_h=img.shape[0], + img_w=600, + threshold=drop_score, + font_path=font_path) + img = np.concatenate([np.array(img), np.array(txt_img)], axis=1) + return img + return image + +def base64_to_cv2(b64str): + import base64 + data = base64.b64decode(b64str.encode('utf8')) + data = np.frombuffer(data, np.uint8) + data = cv2.imdecode(data, cv2.IMREAD_COLOR) + return data + +def str2bool(v): + return v.lower() in ("true", "t", "1") + +def infer_args(): + parser = argparse.ArgumentParser() + # params for prediction engine + parser.add_argument("--use_gpu", type=str2bool, default=True) + parser.add_argument("--use_xpu", type=str2bool, default=False) + parser.add_argument("--use_npu", type=str2bool, default=False) + parser.add_argument("--ir_optim", type=str2bool, default=True) + parser.add_argument("--use_tensorrt", type=str2bool, default=False) + parser.add_argument("--min_subgraph_size", type=int, default=15) + parser.add_argument("--precision", type=str, default="fp32") + parser.add_argument("--gpu_mem", type=int, default=500) + parser.add_argument("--gpu_id", type=int, default=0) + + # params for text detector + parser.add_argument("--image_dir", type=str) + parser.add_argument("--page_num", type=int, default=0) + parser.add_argument("--det_algorithm", type=str, default='DB') + # parser.add_argument("--det_model_dir", type=str, default='./onnx/models/ch_ppocr_server_v2.0/det/det.onnx') + parser.add_argument("--det_model_dir", type=str, default='./pp_onnx/models/ch_PP-OCRv4/ch_PP-OCRv4_det_infer.onnx') + parser.add_argument("--det_limit_side_len", type=float, default=960) + parser.add_argument("--det_limit_type", type=str, default='max') + parser.add_argument("--det_box_type", type=str, default='quad') + + # DB parmas + parser.add_argument("--det_db_thresh", type=float, default=0.3) + parser.add_argument("--det_db_box_thresh", type=float, default=0.6) + parser.add_argument("--det_db_unclip_ratio", type=float, default=1.5) + parser.add_argument("--max_batch_size", type=int, default=10) + parser.add_argument("--use_dilation", type=str2bool, default=False) + parser.add_argument("--det_db_score_mode", type=str, default="fast") + + # # EAST parmas + # parser.add_argument("--det_east_score_thresh", type=float, default=0.8) + # parser.add_argument("--det_east_cover_thresh", type=float, default=0.1) + # parser.add_argument("--det_east_nms_thresh", type=float, default=0.2) + + # # SAST parmas + # parser.add_argument("--det_sast_score_thresh", type=float, default=0.5) + # parser.add_argument("--det_sast_nms_thresh", type=float, default=0.2) + + # # PSE parmas + # parser.add_argument("--det_pse_thresh", type=float, default=0) + # parser.add_argument("--det_pse_box_thresh", type=float, default=0.85) + # parser.add_argument("--det_pse_min_area", type=float, default=16) + # parser.add_argument("--det_pse_scale", type=int, default=1) + + # # FCE parmas + # parser.add_argument("--scales", type=list, default=[8, 16, 32]) + # parser.add_argument("--alpha", type=float, default=1.0) + # parser.add_argument("--beta", type=float, default=1.0) + # parser.add_argument("--fourier_degree", type=int, default=5) + + # params for text recognizer + parser.add_argument("--rec_algorithm", type=str, default='SVTR_LCNet') + # parser.add_argument("--rec_model_dir", type=str, default='./onnx/models/ch_ppocr_server_v2.0/rec/rec.onnx') + parser.add_argument("--rec_model_dir", type=str, default='./pp_onnx/models/ch_PP-OCRv4/ch_PP-OCRv4_rec_infer.onnx') + parser.add_argument("--rec_image_inverse", type=str2bool, default=True) + # parser.add_argument("--rec_image_shape", type=str, default="3, 48, 320") + parser.add_argument("--rec_image_shape", type=str, default="3, 32, 320") + parser.add_argument("--rec_batch_num", type=int, default=6) + parser.add_argument("--max_text_length", type=int, default=25) + parser.add_argument( + "--rec_char_dict_path", + type=str, + default='./pp_onnx/models/ch_ppocr_server_v2.0/ppocr_keys_v1.txt') + parser.add_argument("--use_space_char", type=str2bool, default=True) + parser.add_argument( + "--vis_font_path", type=str, default="./pp_onnx/fonts/simfang.ttf") + parser.add_argument("--drop_score", type=float, default=0.5) + + # params for e2e + parser.add_argument("--e2e_algorithm", type=str, default='PGNet') + parser.add_argument("--e2e_model_dir", type=str) + parser.add_argument("--e2e_limit_side_len", type=float, default=768) + parser.add_argument("--e2e_limit_type", type=str, default='max') + + # PGNet parmas + parser.add_argument("--e2e_pgnet_score_thresh", type=float, default=0.5) + parser.add_argument( + "--e2e_char_dict_path", type=str, default="./onnx/ppocr/utils/ic15_dict.txt") + parser.add_argument("--e2e_pgnet_valid_set", type=str, default='totaltext') + parser.add_argument("--e2e_pgnet_mode", type=str, default='fast') + + # params for text classifier + parser.add_argument("--use_angle_cls", type=str2bool, default=False) + parser.add_argument("--cls_model_dir", type=str, default='./pp_onnx/models/ch_ppocr_server_v2.0/cls/cls.onnx') + parser.add_argument("--cls_image_shape", type=str, default="3, 48, 192") + parser.add_argument("--label_list", type=list, default=['0', '180']) + parser.add_argument("--cls_batch_num", type=int, default=6) + parser.add_argument("--cls_thresh", type=float, default=0.9) + + parser.add_argument("--enable_mkldnn", type=str2bool, default=False) + parser.add_argument("--cpu_threads", type=int, default=10) + parser.add_argument("--use_pdserving", type=str2bool, default=False) + parser.add_argument("--warmup", type=str2bool, default=False) + + # SR parmas + parser.add_argument("--sr_model_dir", type=str) + parser.add_argument("--sr_image_shape", type=str, default="3, 32, 128") + parser.add_argument("--sr_batch_num", type=int, default=1) + + # + parser.add_argument( + "--draw_img_save_dir", type=str, default="./onnx/inference_results") + parser.add_argument("--save_crop_res", type=str2bool, default=False) + parser.add_argument("--crop_res_save_dir", type=str, default="./onnx/output") + + # multi-process + parser.add_argument("--use_mp", type=str2bool, default=False) + parser.add_argument("--total_process_num", type=int, default=1) + parser.add_argument("--process_id", type=int, default=0) + + parser.add_argument("--benchmark", type=str2bool, default=False) + parser.add_argument("--save_log_path", type=str, default="./onnx/log_output/") + + parser.add_argument("--show_log", type=str2bool, default=True) + parser.add_argument("--use_onnx", type=str2bool, default=False) + return parser \ No newline at end of file diff --git a/pp_onnx/logger.py b/pp_onnx/logger.py new file mode 100644 index 0000000..e911c2a --- /dev/null +++ b/pp_onnx/logger.py @@ -0,0 +1,45 @@ +import logging + +LogName = 'Umi-OCR_log' +LogFileName = 'Umi-OCR_debug.log' + + +class Logger: + + def __init__(self): + self.initLogger() + + def initLogger(self): + '''初始化日志''' + + # 日志 + self.logger = logging.getLogger(LogName) + self.logger.setLevel(logging.DEBUG) + + # 控制台 + streamHandler = logging.StreamHandler() + streamHandler.setLevel(logging.DEBUG) + formatPrint = logging.Formatter( + '【%(levelname)s】 %(message)s') + streamHandler.setFormatter(formatPrint) + # self.logger.addHandler(streamHandler) + + return + # 日志文件 + fileHandler = logging.FileHandler(LogFileName) + fileHandler.setLevel(logging.ERROR) + formatFile = logging.Formatter( + ''' +【%(levelname)s】 %(asctime)s +%(message)s + 文件：%(module)s | 函数：%(funcName)s | 行号：%(lineno)d + 线程id：%(thread)d | 线程名：%(thread)s''') + fileHandler.setFormatter(formatFile) + self.logger.addHandler(fileHandler) + + +LOG = Logger() + + +def GetLog(): + return LOG.logger diff --git a/pp_onnx/models/ch_PP-OCRv4/ch_PP-OCRv4_det_infer.onnx b/pp_onnx/models/ch_PP-OCRv4/ch_PP-OCRv4_det_infer.onnx new file mode 100644 index 0000000..64ac2b4 Binary files /dev/null and b/pp_onnx/models/ch_PP-OCRv4/ch_PP-OCRv4_det_infer.onnx differ diff --git a/pp_onnx/models/ch_PP-OCRv4/ch_PP-OCRv4_rec_infer.onnx b/pp_onnx/models/ch_PP-OCRv4/ch_PP-OCRv4_rec_infer.onnx new file mode 100644 index 0000000..bb9ab94 Binary files /dev/null and b/pp_onnx/models/ch_PP-OCRv4/ch_PP-OCRv4_rec_infer.onnx differ diff --git a/pp_onnx/models/ch_PP-OCRv4/ppocrv4_det_rk3588_i8.rknn b/pp_onnx/models/ch_PP-OCRv4/ppocrv4_det_rk3588_i8.rknn new file mode 100644 index 0000000..8298e5d Binary files /dev/null and b/pp_onnx/models/ch_PP-OCRv4/ppocrv4_det_rk3588_i8.rknn differ diff --git a/pp_onnx/models/ch_PP-OCRv4/ppocrv4_rec_rk3588_fp.rknn b/pp_onnx/models/ch_PP-OCRv4/ppocrv4_rec_rk3588_fp.rknn new file mode 100644 index 0000000..33888b3 Binary files /dev/null and b/pp_onnx/models/ch_PP-OCRv4/ppocrv4_rec_rk3588_fp.rknn differ diff --git a/pp_onnx/models/ch_ppocr_server_v2.0/cls/cls.onnx b/pp_onnx/models/ch_ppocr_server_v2.0/cls/cls.onnx new file mode 100644 index 0000000..aab8cdc Binary files /dev/null and b/pp_onnx/models/ch_ppocr_server_v2.0/cls/cls.onnx differ diff --git a/pp_onnx/models/ch_ppocr_server_v2.0/det/det.onnx b/pp_onnx/models/ch_ppocr_server_v2.0/det/det.onnx new file mode 100644 index 0000000..abe1a0d Binary files /dev/null and b/pp_onnx/models/ch_ppocr_server_v2.0/det/det.onnx differ diff --git a/pp_onnx/models/ch_ppocr_server_v2.0/ppocr_keys_v1.txt b/pp_onnx/models/ch_ppocr_server_v2.0/ppocr_keys_v1.txt new file mode 100644 index 0000000..84b885d --- /dev/null +++ b/pp_onnx/models/ch_ppocr_server_v2.0/ppocr_keys_v1.txt @@ -0,0 +1,6623 @@ +' +疗 +绚 +诚 +娇 +溜 +题 +贿 +者 +廖 +更 +纳 +加 +奉 +公 +一 +就 +汴 +计 +与 +路 +房 +原 +妇 +2 +0 +8 +- +7 +其 +> +: +] +, +， +骑 +刈 +全 +消 +昏 +傈 +安 +久 +钟 +嗅 +不 +影 +处 +驽 +蜿 +资 +关 +椤 +地 +瘸 +专 +问 +忖 +票 +嫉 +炎 +韵 +要 +月 +田 +节 +陂 +鄙 +捌 +备 +拳 +伺 +眼 +网 +盎 +大 +傍 +心 +东 +愉 +汇 +蹿 +科 +每 +业 +里 +航 +晏 +字 +平 +录 +先 +1 +3 +彤 +鲶 +产 +稍 +督 +腴 +有 +象 +岳 +注 +绍 +在 +泺 +文 +定 +核 +名 +水 +过 +理 +让 +偷 +率 +等 +这 +发 +” +为 +含 +肥 +酉 +相 +鄱 +七 +编 +猥 +锛 +日 +镀 +蒂 +掰 +倒 +辆 +栾 +栗 +综 +涩 +州 +雌 +滑 +馀 +了 +机 +块 +司 +宰 +甙 +兴 +矽 +抚 +保 +用 +沧 +秩 +如 +收 +息 +滥 +页 +疑 +埠 +! +！ +姥 +异 +橹 +钇 +向 +下 +跄 +的 +椴 +沫 +国 +绥 +獠 +报 +开 +民 +蜇 +何 +分 +凇 +长 +讥 +藏 +掏 +施 +羽 +中 +讲 +派 +嘟 +人 +提 +浼 +间 +世 +而 +古 +多 +倪 +唇 +饯 +控 +庚 +首 +赛 +蜓 +味 +断 +制 +觉 +技 +替 +艰 +溢 +潮 +夕 +钺 +外 +摘 +枋 +动 +双 +单 +啮 +户 +枇 +确 +锦 +曜 +杜 +或 +能 +效 +霜 +盒 +然 +侗 +电 +晁 +放 +步 +鹃 +新 +杖 +蜂 +吒 +濂 +瞬 +评 +总 +隍 +对 +独 +合 +也 +是 +府 +青 +天 +诲 +墙 +组 +滴 +级 +邀 +帘 +示 +已 +时 +骸 +仄 +泅 +和 +遨 +店 +雇 +疫 +持 +巍 +踮 +境 +只 +亨 +目 +鉴 +崤 +闲 +体 +泄 +杂 +作 +般 +轰 +化 +解 +迂 +诿 +蛭 +璀 +腾 +告 +版 +服 +省 +师 +小 +规 +程 +线 +海 +办 +引 +二 +桧 +牌 +砺 +洄 +裴 +修 +图 +痫 +胡 +许 +犊 +事 +郛 +基 +柴 +呼 +食 +研 +奶 +律 +蛋 +因 +葆 +察 +戏 +褒 +戒 +再 +李 +骁 +工 +貂 +油 +鹅 +章 +啄 +休 +场 +给 +睡 +纷 +豆 +器 +捎 +说 +敏 +学 +会 +浒 +设 +诊 +格 +廓 +查 +来 +霓 +室 +溆 +￠ +诡 +寥 +焕 +舜 +柒 +狐 +回 +戟 +砾 +厄 +实 +翩 +尿 +五 +入 +径 +惭 +喹 +股 +宇 +篝 +| +; +美 +期 +云 +九 +祺 +扮 +靠 +锝 +槌 +系 +企 +酰 +阊 +暂 +蚕 +忻 +豁 +本 +羹 +执 +条 +钦 +H +獒 +限 +进 +季 +楦 +于 +芘 +玖 +铋 +茯 +未 +答 +粘 +括 +样 +精 +欠 +矢 +甥 +帷 +嵩 +扣 +令 +仔 +风 +皈 +行 +支 +部 +蓉 +刮 +站 +蜡 +救 +钊 +汗 +松 +嫌 +成 +可 +. +鹤 +院 +从 +交 +政 +怕 +活 +调 +球 +局 +验 +髌 +第 +韫 +谗 +串 +到 +圆 +年 +米 +/ +* +友 +忿 +检 +区 +看 +自 +敢 +刃 +个 +兹 +弄 +流 +留 +同 +没 +齿 +星 +聆 +轼 +湖 +什 +三 +建 +蛔 +儿 +椋 +汕 +震 +颧 +鲤 +跟 +力 +情 +璺 +铨 +陪 +务 +指 +族 +训 +滦 +鄣 +濮 +扒 +商 +箱 +十 +召 +慷 +辗 +所 +莞 +管 +护 +臭 +横 +硒 +嗓 +接 +侦 +六 +露 +党 +馋 +驾 +剖 +高 +侬 +妪 +幂 +猗 +绺 +骐 +央 +酐 +孝 +筝 +课 +徇 +缰 +门 +男 +西 +项 +句 +谙 +瞒 +秃 +篇 +教 +碲 +罚 +声 +呐 +景 +前 +富 +嘴 +鳌 +稀 +免 +朋 +啬 +睐 +去 +赈 +鱼 +住 +肩 +愕 +速 +旁 +波 +厅 +健 +茼 +厥 +鲟 +谅 +投 +攸 +炔 +数 +方 +击 +呋 +谈 +绩 +别 +愫 +僚 +躬 +鹧 +胪 +炳 +招 +喇 +膨 +泵 +蹦 +毛 +结 +5 +4 +谱 +识 +陕 +粽 +婚 +拟 +构 +且 +搜 +任 +潘 +比 +郢 +妨 +醪 +陀 +桔 +碘 +扎 +选 +哈 +骷 +楷 +亿 +明 +缆 +脯 +监 +睫 +逻 +婵 +共 +赴 +淝 +凡 +惦 +及 +达 +揖 +谩 +澹 +减 +焰 +蛹 +番 +祁 +柏 +员 +禄 +怡 +峤 +龙 +白 +叽 +生 +闯 +起 +细 +装 +谕 +竟 +聚 +钙 +上 +导 +渊 +按 +艾 +辘 +挡 +耒 +盹 +饪 +臀 +记 +邮 +蕙 +受 +各 +医 +搂 +普 +滇 +朗 +茸 +带 +翻 +酚 +( +光 +堤 +墟 +蔷 +万 +幻 +〓 +瑙 +辈 +昧 +盏 +亘 +蛀 +吉 +铰 +请 +子 +假 +闻 +税 +井 +诩 +哨 +嫂 +好 +面 +琐 +校 +馊 +鬣 +缂 +营 +访 +炖 +占 +农 +缀 +否 +经 +钚 +棵 +趟 +张 +亟 +吏 +茶 +谨 +捻 +论 +迸 +堂 +玉 +信 +吧 +瞠 +乡 +姬 +寺 +咬 +溏 +苄 +皿 +意 +赉 +宝 +尔 +钰 +艺 +特 +唳 +踉 +都 +荣 +倚 +登 +荐 +丧 +奇 +涵 +批 +炭 +近 +符 +傩 +感 +道 +着 +菊 +虹 +仲 +众 +懈 +濯 +颞 +眺 +南 +释 +北 +缝 +标 +既 +茗 +整 +撼 +迤 +贲 +挎 +耱 +拒 +某 +妍 +卫 +哇 +英 +矶 +藩 +治 +他 +元 +领 +膜 +遮 +穗 +蛾 +飞 +荒 +棺 +劫 +么 +市 +火 +温 +拈 +棚 +洼 +转 +果 +奕 +卸 +迪 +伸 +泳 +斗 +邡 +侄 +涨 +屯 +萋 +胭 +氡 +崮 +枞 +惧 +冒 +彩 +斜 +手 +豚 +随 +旭 +淑 +妞 +形 +菌 +吲 +沱 +争 +驯 +歹 +挟 +兆 +柱 +传 +至 +包 +内 +响 +临 +红 +功 +弩 +衡 +寂 +禁 +老 +棍 +耆 +渍 +织 +害 +氵 +渑 +布 +载 +靥 +嗬 +虽 +苹 +咨 +娄 +库 +雉 +榜 +帜 +嘲 +套 +瑚 +亲 +簸 +欧 +边 +6 +腿 +旮 +抛 +吹 +瞳 +得 +镓 +梗 +厨 +继 +漾 +愣 +憨 +士 +策 +窑 +抑 +躯 +襟 +脏 +参 +贸 +言 +干 +绸 +鳄 +穷 +藜 +音 +折 +详 +) +举 +悍 +甸 +癌 +黎 +谴 +死 +罩 +迁 +寒 +驷 +袖 +媒 +蒋 +掘 +模 +纠 +恣 +观 +祖 +蛆 +碍 +位 +稿 +主 +澧 +跌 +筏 +京 +锏 +帝 +贴 +证 +糠 +才 +黄 +鲸 +略 +炯 +饱 +四 +出 +园 +犀 +牧 +容 +汉 +杆 +浈 +汰 +瑷 +造 +虫 +瘩 +怪 +驴 +济 +应 +花 +沣 +谔 +夙 +旅 +价 +矿 +以 +考 +s +u +呦 +晒 +巡 +茅 +准 +肟 +瓴 +詹 +仟 +褂 +译 +桌 +混 +宁 +怦 +郑 +抿 +些 +余 +鄂 +饴 +攒 +珑 +群 +阖 +岔 +琨 +藓 +预 +环 +洮 +岌 +宀 +杲 +瀵 +最 +常 +囡 +周 +踊 +女 +鼓 +袭 +喉 +简 +范 +薯 +遐 +疏 +粱 +黜 +禧 +法 +箔 +斤 +遥 +汝 +奥 +直 +贞 +撑 +置 +绱 +集 +她 +馅 +逗 +钧 +橱 +魉 +[ +恙 +躁 +唤 +9 +旺 +膘 +待 +脾 +惫 +购 +吗 +依 +盲 +度 +瘿 +蠖 +俾 +之 +镗 +拇 +鲵 +厝 +簧 +续 +款 +展 +啃 +表 +剔 +品 +钻 +腭 +损 +清 +锶 +统 +涌 +寸 +滨 +贪 +链 +吠 +冈 +伎 +迥 +咏 +吁 +览 +防 +迅 +失 +汾 +阔 +逵 +绀 +蔑 +列 +川 +凭 +努 +熨 +揪 +利 +俱 +绉 +抢 +鸨 +我 +即 +责 +膦 +易 +毓 +鹊 +刹 +玷 +岿 +空 +嘞 +绊 +排 +术 +估 +锷 +违 +们 +苟 +铜 +播 +肘 +件 +烫 +审 +鲂 +广 +像 +铌 +惰 +铟 +巳 +胍 +鲍 +康 +憧 +色 +恢 +想 +拷 +尤 +疳 +知 +S +Y +F +D +A +峄 +裕 +帮 +握 +搔 +氐 +氘 +难 +墒 +沮 +雨 +叁 +缥 +悴 +藐 +湫 +娟 +苑 +稠 +颛 +簇 +后 +阕 +闭 +蕤 +缚 +怎 +佞 +码 +嘤 +蔡 +痊 +舱 +螯 +帕 +赫 +昵 +升 +烬 +岫 +、 +疵 +蜻 +髁 +蕨 +隶 +烛 +械 +丑 +盂 +梁 +强 +鲛 +由 +拘 +揉 +劭 +龟 +撤 +钩 +呕 +孛 +费 +妻 +漂 +求 +阑 +崖 +秤 +甘 +通 +深 +补 +赃 +坎 +床 +啪 +承 +吼 +量 +暇 +钼 +烨 +阂 +擎 +脱 +逮 +称 +P +神 +属 +矗 +华 +届 +狍 +葑 +汹 +育 +患 +窒 +蛰 +佼 +静 +槎 +运 +鳗 +庆 +逝 +曼 +疱 +克 +代 +官 +此 +麸 +耧 +蚌 +晟 +例 +础 +榛 +副 +测 +唰 +缢 +迹 +灬 +霁 +身 +岁 +赭 +扛 +又 +菡 +乜 +雾 +板 +读 +陷 +徉 +贯 +郁 +虑 +变 +钓 +菜 +圾 +现 +琢 +式 +乐 +维 +渔 +浜 +左 +吾 +脑 +钡 +警 +T +啵 +拴 +偌 +漱 +湿 +硕 +止 +骼 +魄 +积 +燥 +联 +踢 +玛 +则 +窿 +见 +振 +畿 +送 +班 +钽 +您 +赵 +刨 +印 +讨 +踝 +籍 +谡 +舌 +崧 +汽 +蔽 +沪 +酥 +绒 +怖 +财 +帖 +肱 +私 +莎 +勋 +羔 +霸 +励 +哼 +帐 +将 +帅 +渠 +纪 +婴 +娩 +岭 +厘 +滕 +吻 +伤 +坝 +冠 +戊 +隆 +瘁 +介 +涧 +物 +黍 +并 +姗 +奢 +蹑 +掣 +垸 +锴 +命 +箍 +捉 +病 +辖 +琰 +眭 +迩 +艘 +绌 +繁 +寅 +若 +毋 +思 +诉 +类 +诈 +燮 +轲 +酮 +狂 +重 +反 +职 +筱 +县 +委 +磕 +绣 +奖 +晋 +濉 +志 +徽 +肠 +呈 +獐 +坻 +口 +片 +碰 +几 +村 +柿 +劳 +料 +获 +亩 +惕 +晕 +厌 +号 +罢 +池 +正 +鏖 +煨 +家 +棕 +复 +尝 +懋 +蜥 +锅 +岛 +扰 +队 +坠 +瘾 +钬 +@ +卧 +疣 +镇 +譬 +冰 +彷 +频 +黯 +据 +垄 +采 +八 +缪 +瘫 +型 +熹 +砰 +楠 +襁 +箐 +但 +嘶 +绳 +啤 +拍 +盥 +穆 +傲 +洗 +盯 +塘 +怔 +筛 +丿 +台 +恒 +喂 +葛 +永 +￥ +烟 +酒 +桦 +书 +砂 +蚝 +缉 +态 +瀚 +袄 +圳 +轻 +蛛 +超 +榧 +遛 +姒 +奘 +铮 +右 +荽 +望 +偻 +卡 +丶 +氰 +附 +做 +革 +索 +戚 +坨 +桷 +唁 +垅 +榻 +岐 +偎 +坛 +莨 +山 +殊 +微 +骇 +陈 +爨 +推 +嗝 +驹 +澡 +藁 +呤 +卤 +嘻 +糅 +逛 +侵 +郓 +酌 +德 +摇 +※ +鬃 +被 +慨 +殡 +羸 +昌 +泡 +戛 +鞋 +河 +宪 +沿 +玲 +鲨 +翅 +哽 +源 +铅 +语 +照 +邯 +址 +荃 +佬 +顺 +鸳 +町 +霭 +睾 +瓢 +夸 +椁 +晓 +酿 +痈 +咔 +侏 +券 +噎 +湍 +签 +嚷 +离 +午 +尚 +社 +锤 +背 +孟 +使 +浪 +缦 +潍 +鞅 +军 +姹 +驶 +笑 +鳟 +鲁 +》 +孽 +钜 +绿 +洱 +礴 +焯 +椰 +颖 +囔 +乌 +孔 +巴 +互 +性 +椽 +哞 +聘 +昨 +早 +暮 +胶 +炀 +隧 +低 +彗 +昝 +铁 +呓 +氽 +藉 +喔 +癖 +瑗 +姨 +权 +胱 +韦 +堑 +蜜 +酋 +楝 +砝 +毁 +靓 +歙 +锲 +究 +屋 +喳 +骨 +辨 +碑 +武 +鸠 +宫 +辜 +烊 +适 +坡 +殃 +培 +佩 +供 +走 +蜈 +迟 +翼 +况 +姣 +凛 +浔 +吃 +飘 +债 +犟 +金 +促 +苛 +崇 +坂 +莳 +畔 +绂 +兵 +蠕 +斋 +根 +砍 +亢 +欢 +恬 +崔 +剁 +餐 +榫 +快 +扶 +‖ +濒 +缠 +鳜 +当 +彭 +驭 +浦 +篮 +昀 +锆 +秸 +钳 +弋 +娣 +瞑 +夷 +龛 +苫 +拱 +致 +% +嵊 +障 +隐 +弑 +初 +娓 +抉 +汩 +累 +蓖 +" +唬 +助 +苓 +昙 +押 +毙 +破 +城 +郧 +逢 +嚏 +獭 +瞻 +溱 +婿 +赊 +跨 +恼 +璧 +萃 +姻 +貉 +灵 +炉 +密 +氛 +陶 +砸 +谬 +衔 +点 +琛 +沛 +枳 +层 +岱 +诺 +脍 +榈 +埂 +征 +冷 +裁 +打 +蹴 +素 +瘘 +逞 +蛐 +聊 +激 +腱 +萘 +踵 +飒 +蓟 +吆 +取 +咙 +簋 +涓 +矩 +曝 +挺 +揣 +座 +你 +史 +舵 +焱 +尘 +苏 +笈 +脚 +溉 +榨 +诵 +樊 +邓 +焊 +义 +庶 +儋 +蟋 +蒲 +赦 +呷 +杞 +诠 +豪 +还 +试 +颓 +茉 +太 +除 +紫 +逃 +痴 +草 +充 +鳕 +珉 +祗 +墨 +渭 +烩 +蘸 +慕 +璇 +镶 +穴 +嵘 +恶 +骂 +险 +绋 +幕 +碉 +肺 +戳 +刘 +潞 +秣 +纾 +潜 +銮 +洛 +须 +罘 +销 +瘪 +汞 +兮 +屉 +r +林 +厕 +质 +探 +划 +狸 +殚 +善 +煊 +烹 +〒 +锈 +逯 +宸 +辍 +泱 +柚 +袍 +远 +蹋 +嶙 +绝 +峥 +娥 +缍 +雀 +徵 +认 +镱 +谷 += +贩 +勉 +撩 +鄯 +斐 +洋 +非 +祚 +泾 +诒 +饿 +撬 +威 +晷 +搭 +芍 +锥 +笺 +蓦 +候 +琊 +档 +礁 +沼 +卵 +荠 +忑 +朝 +凹 +瑞 +头 +仪 +弧 +孵 +畏 +铆 +突 +衲 +车 +浩 +气 +茂 +悖 +厢 +枕 +酝 +戴 +湾 +邹 +飚 +攘 +锂 +写 +宵 +翁 +岷 +无 +喜 +丈 +挑 +嗟 +绛 +殉 +议 +槽 +具 +醇 +淞 +笃 +郴 +阅 +饼 +底 +壕 +砚 +弈 +询 +缕 +庹 +翟 +零 +筷 +暨 +舟 +闺 +甯 +撞 +麂 +茌 +蔼 +很 +珲 +捕 +棠 +角 +阉 +媛 +娲 +诽 +剿 +尉 +爵 +睬 +韩 +诰 +匣 +危 +糍 +镯 +立 +浏 +阳 +少 +盆 +舔 +擘 +匪 +申 +尬 +铣 +旯 +抖 +赘 +瓯 +居 +ˇ +哮 +游 +锭 +茏 +歌 +坏 +甚 +秒 +舞 +沙 +仗 +劲 +潺 +阿 +燧 +郭 +嗖 +霏 +忠 +材 +奂 +耐 +跺 +砀 +输 +岖 +媳 +氟 +极 +摆 +灿 +今 +扔 +腻 +枝 +奎 +药 +熄 +吨 +话 +q +额 +慑 +嘌 +协 +喀 +壳 +埭 +视 +著 +於 +愧 +陲 +翌 +峁 +颅 +佛 +腹 +聋 +侯 +咎 +叟 +秀 +颇 +存 +较 +罪 +哄 +岗 +扫 +栏 +钾 +羌 +己 +璨 +枭 +霉 +煌 +涸 +衿 +键 +镝 +益 +岢 +奏 +连 +夯 +睿 +冥 +均 +糖 +狞 +蹊 +稻 +爸 +刿 +胥 +煜 +丽 +肿 +璃 +掸 +跚 +灾 +垂 +樾 +濑 +乎 +莲 +窄 +犹 +撮 +战 +馄 +软 +络 +显 +鸢 +胸 +宾 +妲 +恕 +埔 +蝌 +份 +遇 +巧 +瞟 +粒 +恰 +剥 +桡 +博 +讯 +凯 +堇 +阶 +滤 +卖 +斌 +骚 +彬 +兑 +磺 +樱 +舷 +两 +娱 +福 +仃 +差 +找 +桁 +÷ +净 +把 +阴 +污 +戬 +雷 +碓 +蕲 +楚 +罡 +焖 +抽 +妫 +咒 +仑 +闱 +尽 +邑 +菁 +爱 +贷 +沥 +鞑 +牡 +嗉 +崴 +骤 +塌 +嗦 +订 +拮 +滓 +捡 +锻 +次 +坪 +杩 +臃 +箬 +融 +珂 +鹗 +宗 +枚 +降 +鸬 +妯 +阄 +堰 +盐 +毅 +必 +杨 +崃 +俺 +甬 +状 +莘 +货 +耸 +菱 +腼 +铸 +唏 +痤 +孚 +澳 +懒 +溅 +翘 +疙 +杷 +淼 +缙 +骰 +喊 +悉 +砻 +坷 +艇 +赁 +界 +谤 +纣 +宴 +晃 +茹 +归 +饭 +梢 +铡 +街 +抄 +肼 +鬟 +苯 +颂 +撷 +戈 +炒 +咆 +茭 +瘙 +负 +仰 +客 +琉 +铢 +封 +卑 +珥 +椿 +镧 +窨 +鬲 +寿 +御 +袤 +铃 +萎 +砖 +餮 +脒 +裳 +肪 +孕 +嫣 +馗 +嵇 +恳 +氯 +江 +石 +褶 +冢 +祸 +阻 +狈 +羞 +银 +靳 +透 +咳 +叼 +敷 +芷 +啥 +它 +瓤 +兰 +痘 +懊 +逑 +肌 +往 +捺 +坊 +甩 +呻 +〃 +沦 +忘 +膻 +祟 +菅 +剧 +崆 +智 +坯 +臧 +霍 +墅 +攻 +眯 +倘 +拢 +骠 +铐 +庭 +岙 +瓠 +′ +缺 +泥 +迢 +捶 +? +？ +郏 +喙 +掷 +沌 +纯 +秘 +种 +听 +绘 +固 +螨 +团 +香 +盗 +妒 +埚 +蓝 +拖 +旱 +荞 +铀 +血 +遏 +汲 +辰 +叩 +拽 +幅 +硬 +惶 +桀 +漠 +措 +泼 +唑 +齐 +肾 +念 +酱 +虚 +屁 +耶 +旗 +砦 +闵 +婉 +馆 +拭 +绅 +韧 +忏 +窝 +醋 +葺 +顾 +辞 +倜 +堆 +辋 +逆 +玟 +贱 +疾 +董 +惘 +倌 +锕 +淘 +嘀 +莽 +俭 +笏 +绑 +鲷 +杈 +择 +蟀 +粥 +嗯 +驰 +逾 +案 +谪 +褓 +胫 +哩 +昕 +颚 +鲢 +绠 +躺 +鹄 +崂 +儒 +俨 +丝 +尕 +泌 +啊 +萸 +彰 +幺 +吟 +骄 +苣 +弦 +脊 +瑰 +〈 +诛 +镁 +析 +闪 +剪 +侧 +哟 +框 +螃 +守 +嬗 +燕 +狭 +铈 +缮 +概 +迳 +痧 +鲲 +俯 +售 +笼 +痣 +扉 +挖 +满 +咋 +援 +邱 +扇 +歪 +便 +玑 +绦 +峡 +蛇 +叨 +〖 +泽 +胃 +斓 +喋 +怂 +坟 +猪 +该 +蚬 +炕 +弥 +赞 +棣 +晔 +娠 +挲 +狡 +创 +疖 +铕 +镭 +稷 +挫 +弭 +啾 +翔 +粉 +履 +苘 +哦 +楼 +秕 +铂 +土 +锣 +瘟 +挣 +栉 +习 +享 +桢 +袅 +磨 +桂 +谦 +延 +坚 +蔚 +噗 +署 +谟 +猬 +钎 +恐 +嬉 +雒 +倦 +衅 +亏 +璩 +睹 +刻 +殿 +王 +算 +雕 +麻 +丘 +柯 +骆 +丸 +塍 +谚 +添 +鲈 +垓 +桎 +蚯 +芥 +予 +飕 +镦 +谌 +窗 +醚 +菀 +亮 +搪 +莺 +蒿 +羁 +足 +J +真 +轶 +悬 +衷 +靛 +翊 +掩 +哒 +炅 +掐 +冼 +妮 +l +谐 +稚 +荆 +擒 +犯 +陵 +虏 +浓 +崽 +刍 +陌 +傻 +孜 +千 +靖 +演 +矜 +钕 +煽 +杰 +酗 +渗 +伞 +栋 +俗 +泫 +戍 +罕 +沾 +疽 +灏 +煦 +芬 +磴 +叱 +阱 +榉 +湃 +蜀 +叉 +醒 +彪 +租 +郡 +篷 +屎 +良 +垢 +隗 +弱 +陨 +峪 +砷 +掴 +颁 +胎 +雯 +绵 +贬 +沐 +撵 +隘 +篙 +暖 +曹 +陡 +栓 +填 +臼 +彦 +瓶 +琪 +潼 +哪 +鸡 +摩 +啦 +俟 +锋 +域 +耻 +蔫 +疯 +纹 +撇 +毒 +绶 +痛 +酯 +忍 +爪 +赳 +歆 +嘹 +辕 +烈 +册 +朴 +钱 +吮 +毯 +癜 +娃 +谀 +邵 +厮 +炽 +璞 +邃 +丐 +追 +词 +瓒 +忆 +轧 +芫 +谯 +喷 +弟 +半 +冕 +裙 +掖 +墉 +绮 +寝 +苔 +势 +顷 +褥 +切 +衮 +君 +佳 +嫒 +蚩 +霞 +佚 +洙 +逊 +镖 +暹 +唛 +& +殒 +顶 +碗 +獗 +轭 +铺 +蛊 +废 +恹 +汨 +崩 +珍 +那 +杵 +曲 +纺 +夏 +薰 +傀 +闳 +淬 +姘 +舀 +拧 +卷 +楂 +恍 +讪 +厩 +寮 +篪 +赓 +乘 +灭 +盅 +鞣 +沟 +慎 +挂 +饺 +鼾 +杳 +树 +缨 +丛 +絮 +娌 +臻 +嗳 +篡 +侩 +述 +衰 +矛 +圈 +蚜 +匕 +筹 +匿 +濞 +晨 +叶 +骋 +郝 +挚 +蚴 +滞 +增 +侍 +描 +瓣 +吖 +嫦 +蟒 +匾 +圣 +赌 +毡 +癞 +恺 +百 +曳 +需 +篓 +肮 +庖 +帏 +卿 +驿 +遗 +蹬 +鬓 +骡 +歉 +芎 +胳 +屐 +禽 +烦 +晌 +寄 +媾 +狄 +翡 +苒 +船 +廉 +终 +痞 +殇 +々 +畦 +饶 +改 +拆 +悻 +萄 +￡ +瓿 +乃 +訾 +桅 +匮 +溧 +拥 +纱 +铍 +骗 +蕃 +龋 +缬 +父 +佐 +疚 +栎 +醍 +掳 +蓄 +x +惆 +颜 +鲆 +榆 +〔 +猎 +敌 +暴 +谥 +鲫 +贾 +罗 +玻 +缄 +扦 +芪 +癣 +落 +徒 +臾 +恿 +猩 +托 +邴 +肄 +牵 +春 +陛 +耀 +刊 +拓 +蓓 +邳 +堕 +寇 +枉 +淌 +啡 +湄 +兽 +酷 +萼 +碚 +濠 +萤 +夹 +旬 +戮 +梭 +琥 +椭 +昔 +勺 +蜊 +绐 +晚 +孺 +僵 +宣 +摄 +冽 +旨 +萌 +忙 +蚤 +眉 +噼 +蟑 +付 +契 +瓜 +悼 +颡 +壁 +曾 +窕 +颢 +澎 +仿 +俑 +浑 +嵌 +浣 +乍 +碌 +褪 +乱 +蔟 +隙 +玩 +剐 +葫 +箫 +纲 +围 +伐 +决 +伙 +漩 +瑟 +刑 +肓 +镳 +缓 +蹭 +氨 +皓 +典 +畲 +坍 +铑 +檐 +塑 +洞 +倬 +储 +胴 +淳 +戾 +吐 +灼 +惺 +妙 +毕 +珐 +缈 +虱 +盖 +羰 +鸿 +磅 +谓 +髅 +娴 +苴 +唷 +蚣 +霹 +抨 +贤 +唠 +犬 +誓 +逍 +庠 +逼 +麓 +籼 +釉 +呜 +碧 +秧 +氩 +摔 +霄 +穸 +纨 +辟 +妈 +映 +完 +牛 +缴 +嗷 +炊 +恩 +荔 +茆 +掉 +紊 +慌 +莓 +羟 +阙 +萁 +磐 +另 +蕹 +辱 +鳐 +湮 +吡 +吩 +唐 +睦 +垠 +舒 +圜 +冗 +瞿 +溺 +芾 +囱 +匠 +僳 +汐 +菩 +饬 +漓 +黑 +霰 +浸 +濡 +窥 +毂 +蒡 +兢 +驻 +鹉 +芮 +诙 +迫 +雳 +厂 +忐 +臆 +猴 +鸣 +蚪 +栈 +箕 +羡 +渐 +莆 +捍 +眈 +哓 +趴 +蹼 +埕 +嚣 +骛 +宏 +淄 +斑 +噜 +严 +瑛 +垃 +椎 +诱 +压 +庾 +绞 +焘 +廿 +抡 +迄 +棘 +夫 +纬 +锹 +眨 +瞌 +侠 +脐 +竞 +瀑 +孳 +骧 +遁 +姜 +颦 +荪 +滚 +萦 +伪 +逸 +粳 +爬 +锁 +矣 +役 +趣 +洒 +颔 +诏 +逐 +奸 +甭 +惠 +攀 +蹄 +泛 +尼 +拼 +阮 +鹰 +亚 +颈 +惑 +勒 +〉 +际 +肛 +爷 +刚 +钨 +丰 +养 +冶 +鲽 +辉 +蔻 +画 +覆 +皴 +妊 +麦 +返 +醉 +皂 +擀 +〗 +酶 +凑 +粹 +悟 +诀 +硖 +港 +卜 +z +杀 +涕 +± +舍 +铠 +抵 +弛 +段 +敝 +镐 +奠 +拂 +轴 +跛 +袱 +e +t +沉 +菇 +俎 +薪 +峦 +秭 +蟹 +历 +盟 +菠 +寡 +液 +肢 +喻 +染 +裱 +悱 +抱 +氙 +赤 +捅 +猛 +跑 +氮 +谣 +仁 +尺 +辊 +窍 +烙 +衍 +架 +擦 +倏 +璐 +瑁 +币 +楞 +胖 +夔 +趸 +邛 +惴 +饕 +虔 +蝎 +§ +哉 +贝 +宽 +辫 +炮 +扩 +饲 +籽 +魏 +菟 +锰 +伍 +猝 +末 +琳 +哚 +蛎 +邂 +呀 +姿 +鄞 +却 +歧 +仙 +恸 +椐 +森 +牒 +寤 +袒 +婆 +虢 +雅 +钉 +朵 +贼 +欲 +苞 +寰 +故 +龚 +坭 +嘘 +咫 +礼 +硷 +兀 +睢 +汶 +’ +铲 +烧 +绕 +诃 +浃 +钿 +哺 +柜 +讼 +颊 +璁 +腔 +洽 +咐 +脲 +簌 +筠 +镣 +玮 +鞠 +谁 +兼 +姆 +挥 +梯 +蝴 +谘 +漕 +刷 +躏 +宦 +弼 +b +垌 +劈 +麟 +莉 +揭 +笙 +渎 +仕 +嗤 +仓 +配 +怏 +抬 +错 +泯 +镊 +孰 +猿 +邪 +仍 +秋 +鼬 +壹 +歇 +吵 +炼 +< +尧 +射 +柬 +廷 +胧 +霾 +凳 +隋 +肚 +浮 +梦 +祥 +株 +堵 +退 +L +鹫 +跎 +凶 +毽 +荟 +炫 +栩 +玳 +甜 +沂 +鹿 +顽 +伯 +爹 +赔 +蛴 +徐 +匡 +欣 +狰 +缸 +雹 +蟆 +疤 +默 +沤 +啜 +痂 +衣 +禅 +w +i +h +辽 +葳 +黝 +钗 +停 +沽 +棒 +馨 +颌 +肉 +吴 +硫 +悯 +劾 +娈 +马 +啧 +吊 +悌 +镑 +峭 +帆 +瀣 +涉 +咸 +疸 +滋 +泣 +翦 +拙 +癸 +钥 +蜒 ++ +尾 +庄 +凝 +泉 +婢 +渴 +谊 +乞 +陆 +锉 +糊 +鸦 +淮 +I +B +N +晦 +弗 +乔 +庥 +葡 +尻 +席 +橡 +傣 +渣 +拿 +惩 +麋 +斛 +缃 +矮 +蛏 +岘 +鸽 +姐 +膏 +催 +奔 +镒 +喱 +蠡 +摧 +钯 +胤 +柠 +拐 +璋 +鸥 +卢 +荡 +倾 +^ +_ +珀 +逄 +萧 +塾 +掇 +贮 +笆 +聂 +圃 +冲 +嵬 +M +滔 +笕 +值 +炙 +偶 +蜱 +搐 +梆 +汪 +蔬 +腑 +鸯 +蹇 +敞 +绯 +仨 +祯 +谆 +梧 +糗 +鑫 +啸 +豺 +囹 +猾 +巢 +柄 +瀛 +筑 +踌 +沭 +暗 +苁 +鱿 +蹉 +脂 +蘖 +牢 +热 +木 +吸 +溃 +宠 +序 +泞 +偿 +拜 +檩 +厚 +朐 +毗 +螳 +吞 +媚 +朽 +担 +蝗 +橘 +畴 +祈 +糟 +盱 +隼 +郜 +惜 +珠 +裨 +铵 +焙 +琚 +唯 +咚 +噪 +骊 +丫 +滢 +勤 +棉 +呸 +咣 +淀 +隔 +蕾 +窈 +饨 +挨 +煅 +短 +匙 +粕 +镜 +赣 +撕 +墩 +酬 +馁 +豌 +颐 +抗 +酣 +氓 +佑 +搁 +哭 +递 +耷 +涡 +桃 +贻 +碣 +截 +瘦 +昭 +镌 +蔓 +氚 +甲 +猕 +蕴 +蓬 +散 +拾 +纛 +狼 +猷 +铎 +埋 +旖 +矾 +讳 +囊 +糜 +迈 +粟 +蚂 +紧 +鲳 +瘢 +栽 +稼 +羊 +锄 +斟 +睁 +桥 +瓮 +蹙 +祉 +醺 +鼻 +昱 +剃 +跳 +篱 +跷 +蒜 +翎 +宅 +晖 +嗑 +壑 +峻 +癫 +屏 +狠 +陋 +袜 +途 +憎 +祀 +莹 +滟 +佶 +溥 +臣 +约 +盛 +峰 +磁 +慵 +婪 +拦 +莅 +朕 +鹦 +粲 +裤 +哎 +疡 +嫖 +琵 +窟 +堪 +谛 +嘉 +儡 +鳝 +斩 +郾 +驸 +酊 +妄 +胜 +贺 +徙 +傅 +噌 +钢 +栅 +庇 +恋 +匝 +巯 +邈 +尸 +锚 +粗 +佟 +蛟 +薹 +纵 +蚊 +郅 +绢 +锐 +苗 +俞 +篆 +淆 +膀 +鲜 +煎 +诶 +秽 +寻 +涮 +刺 +怀 +噶 +巨 +褰 +魅 +灶 +灌 +桉 +藕 +谜 +舸 +薄 +搀 +恽 +借 +牯 +痉 +渥 +愿 +亓 +耘 +杠 +柩 +锔 +蚶 +钣 +珈 +喘 +蹒 +幽 +赐 +稗 +晤 +莱 +泔 +扯 +肯 +菪 +裆 +腩 +豉 +疆 +骜 +腐 +倭 +珏 +唔 +粮 +亡 +润 +慰 +伽 +橄 +玄 +誉 +醐 +胆 +龊 +粼 +塬 +陇 +彼 +削 +嗣 +绾 +芽 +妗 +垭 +瘴 +爽 +薏 +寨 +龈 +泠 +弹 +赢 +漪 +猫 +嘧 +涂 +恤 +圭 +茧 +烽 +屑 +痕 +巾 +赖 +荸 +凰 +腮 +畈 +亵 +蹲 +偃 +苇 +澜 +艮 +换 +骺 +烘 +苕 +梓 +颉 +肇 +哗 +悄 +氤 +涠 +葬 +屠 +鹭 +植 +竺 +佯 +诣 +鲇 +瘀 +鲅 +邦 +移 +滁 +冯 +耕 +癔 +戌 +茬 +沁 +巩 +悠 +湘 +洪 +痹 +锟 +循 +谋 +腕 +鳃 +钠 +捞 +焉 +迎 +碱 +伫 +急 +榷 +奈 +邝 +卯 +辄 +皲 +卟 +醛 +畹 +忧 +稳 +雄 +昼 +缩 +阈 +睑 +扌 +耗 +曦 +涅 +捏 +瞧 +邕 +淖 +漉 +铝 +耦 +禹 +湛 +喽 +莼 +琅 +诸 +苎 +纂 +硅 +始 +嗨 +傥 +燃 +臂 +赅 +嘈 +呆 +贵 +屹 +壮 +肋 +亍 +蚀 +卅 +豹 +腆 +邬 +迭 +浊 +} +童 +螂 +捐 +圩 +勐 +触 +寞 +汊 +壤 +荫 +膺 +渌 +芳 +懿 +遴 +螈 +泰 +蓼 +蛤 +茜 +舅 +枫 +朔 +膝 +眙 +避 +梅 +判 +鹜 +璜 +牍 +缅 +垫 +藻 +黔 +侥 +惚 +懂 +踩 +腰 +腈 +札 +丞 +唾 +慈 +顿 +摹 +荻 +琬 +~ +斧 +沈 +滂 +胁 +胀 +幄 +莜 +Z +匀 +鄄 +掌 +绰 +茎 +焚 +赋 +萱 +谑 +汁 +铒 +瞎 +夺 +蜗 +野 +娆 +冀 +弯 +篁 +懵 +灞 +隽 +芡 +脘 +俐 +辩 +芯 +掺 +喏 +膈 +蝈 +觐 +悚 +踹 +蔗 +熠 +鼠 +呵 +抓 +橼 +峨 +畜 +缔 +禾 +崭 +弃 +熊 +摒 +凸 +拗 +穹 +蒙 +抒 +祛 +劝 +闫 +扳 +阵 +醌 +踪 +喵 +侣 +搬 +仅 +荧 +赎 +蝾 +琦 +买 +婧 +瞄 +寓 +皎 +冻 +赝 +箩 +莫 +瞰 +郊 +笫 +姝 +筒 +枪 +遣 +煸 +袋 +舆 +痱 +涛 +母 +〇 +启 +践 +耙 +绲 +盘 +遂 +昊 +搞 +槿 +诬 +纰 +泓 +惨 +檬 +亻 +越 +C +o +憩 +熵 +祷 +钒 +暧 +塔 +阗 +胰 +咄 +娶 +魔 +琶 +钞 +邻 +扬 +杉 +殴 +咽 +弓 +〆 +髻 +】 +吭 +揽 +霆 +拄 +殖 +脆 +彻 +岩 +芝 +勃 +辣 +剌 +钝 +嘎 +甄 +佘 +皖 +伦 +授 +徕 +憔 +挪 +皇 +庞 +稔 +芜 +踏 +溴 +兖 +卒 +擢 +饥 +鳞 +煲 +‰ +账 +颗 +叻 +斯 +捧 +鳍 +琮 +讹 +蛙 +纽 +谭 +酸 +兔 +莒 +睇 +伟 +觑 +羲 +嗜 +宜 +褐 +旎 +辛 +卦 +诘 +筋 +鎏 +溪 +挛 +熔 +阜 +晰 +鳅 +丢 +奚 +灸 +呱 +献 +陉 +黛 +鸪 +甾 +萨 +疮 +拯 +洲 +疹 +辑 +叙 +恻 +谒 +允 +柔 +烂 +氏 +逅 +漆 +拎 +惋 +扈 +湟 +纭 +啕 +掬 +擞 +哥 +忽 +涤 +鸵 +靡 +郗 +瓷 +扁 +廊 +怨 +雏 +钮 +敦 +E +懦 +憋 +汀 +拚 +啉 +腌 +岸 +f +痼 +瞅 +尊 +咀 +眩 +飙 +忌 +仝 +迦 +熬 +毫 +胯 +篑 +茄 +腺 +凄 +舛 +碴 +锵 +诧 +羯 +後 +漏 +汤 +宓 +仞 +蚁 +壶 +谰 +皑 +铄 +棰 +罔 +辅 +晶 +苦 +牟 +闽 +\ +烃 +饮 +聿 +丙 +蛳 +朱 +煤 +涔 +鳖 +犁 +罐 +荼 +砒 +淦 +妤 +黏 +戎 +孑 +婕 +瑾 +戢 +钵 +枣 +捋 +砥 +衩 +狙 +桠 +稣 +阎 +肃 +梏 +诫 +孪 +昶 +婊 +衫 +嗔 +侃 +塞 +蜃 +樵 +峒 +貌 +屿 +欺 +缫 +阐 +栖 +诟 +珞 +荭 +吝 +萍 +嗽 +恂 +啻 +蜴 +磬 +峋 +俸 +豫 +谎 +徊 +镍 +韬 +魇 +晴 +U +囟 +猜 +蛮 +坐 +囿 +伴 +亭 +肝 +佗 +蝠 +妃 +胞 +滩 +榴 +氖 +垩 +苋 +砣 +扪 +馏 +姓 +轩 +厉 +夥 +侈 +禀 +垒 +岑 +赏 +钛 +辐 +痔 +披 +纸 +碳 +“ +坞 +蠓 +挤 +荥 +沅 +悔 +铧 +帼 +蒌 +蝇 +a +p +y +n +g +哀 +浆 +瑶 +凿 +桶 +馈 +皮 +奴 +苜 +佤 +伶 +晗 +铱 +炬 +优 +弊 +氢 +恃 +甫 +攥 +端 +锌 +灰 +稹 +炝 +曙 +邋 +亥 +眶 +碾 +拉 +萝 +绔 +捷 +浍 +腋 +姑 +菖 +凌 +涞 +麽 +锢 +桨 +潢 +绎 +镰 +殆 +锑 +渝 +铬 +困 +绽 +觎 +匈 +糙 +暑 +裹 +鸟 +盔 +肽 +迷 +綦 +『 +亳 +佝 +俘 +钴 +觇 +骥 +仆 +疝 +跪 +婶 +郯 +瀹 +唉 +脖 +踞 +针 +晾 +忒 +扼 +瞩 +叛 +椒 +疟 +嗡 +邗 +肆 +跆 +玫 +忡 +捣 +咧 +唆 +艄 +蘑 +潦 +笛 +阚 +沸 +泻 +掊 +菽 +贫 +斥 +髂 +孢 +镂 +赂 +麝 +鸾 +屡 +衬 +苷 +恪 +叠 +希 +粤 +爻 +喝 +茫 +惬 +郸 +绻 +庸 +撅 +碟 +宄 +妹 +膛 +叮 +饵 +崛 +嗲 +椅 +冤 +搅 +咕 +敛 +尹 +垦 +闷 +蝉 +霎 +勰 +败 +蓑 +泸 +肤 +鹌 +幌 +焦 +浠 +鞍 +刁 +舰 +乙 +竿 +裔 +。 +茵 +函 +伊 +兄 +丨 +娜 +匍 +謇 +莪 +宥 +似 +蝽 +翳 +酪 +翠 +粑 +薇 +祢 +骏 +赠 +叫 +Q +噤 +噻 +竖 +芗 +莠 +潭 +俊 +羿 +耜 +O +郫 +趁 +嗪 +囚 +蹶 +芒 +洁 +笋 +鹑 +敲 +硝 +啶 +堡 +渲 +揩 +』 +携 +宿 +遒 +颍 +扭 +棱 +割 +萜 +蔸 +葵 +琴 +捂 +饰 +衙 +耿 +掠 +募 +岂 +窖 +涟 +蔺 +瘤 +柞 +瞪 +怜 +匹 +距 +楔 +炜 +哆 +秦 +缎 +幼 +茁 +绪 +痨 +恨 +楸 +娅 +瓦 +桩 +雪 +嬴 +伏 +榔 +妥 +铿 +拌 +眠 +雍 +缇 +‘ +卓 +搓 +哌 +觞 +噩 +屈 +哧 +髓 +咦 +巅 +娑 +侑 +淫 +膳 +祝 +勾 +姊 +莴 +胄 +疃 +薛 +蜷 +胛 +巷 +芙 +芋 +熙 +闰 +勿 +窃 +狱 +剩 +钏 +幢 +陟 +铛 +慧 +靴 +耍 +k +浙 +浇 +飨 +惟 +绗 +祜 +澈 +啼 +咪 +磷 +摞 +诅 +郦 +抹 +跃 +壬 +吕 +肖 +琏 +颤 +尴 +剡 +抠 +凋 +赚 +泊 +津 +宕 +殷 +倔 +氲 +漫 +邺 +涎 +怠 +$ +垮 +荬 +遵 +俏 +叹 +噢 +饽 +蜘 +孙 +筵 +疼 +鞭 +羧 +牦 +箭 +潴 +c +眸 +祭 +髯 +啖 +坳 +愁 +芩 +驮 +倡 +巽 +穰 +沃 +胚 +怒 +凤 +槛 +剂 +趵 +嫁 +v +邢 +灯 +鄢 +桐 +睽 +檗 +锯 +槟 +婷 +嵋 +圻 +诗 +蕈 +颠 +遭 +痢 +芸 +怯 +馥 +竭 +锗 +徜 +恭 +遍 +籁 +剑 +嘱 +苡 +龄 +僧 +桑 +潸 +弘 +澶 +楹 +悲 +讫 +愤 +腥 +悸 +谍 +椹 +呢 +桓 +葭 +攫 +阀 +翰 +躲 +敖 +柑 +郎 +笨 +橇 +呃 +魁 +燎 +脓 +葩 +磋 +垛 +玺 +狮 +沓 +砜 +蕊 +锺 +罹 +蕉 +翱 +虐 +闾 +巫 +旦 +茱 +嬷 +枯 +鹏 +贡 +芹 +汛 +矫 +绁 +拣 +禺 +佃 +讣 +舫 +惯 +乳 +趋 +疲 +挽 +岚 +虾 +衾 +蠹 +蹂 +飓 +氦 +铖 +孩 +稞 +瑜 +壅 +掀 +勘 +妓 +畅 +髋 +W +庐 +牲 +蓿 +榕 +练 +垣 +唱 +邸 +菲 +昆 +婺 +穿 +绡 +麒 +蚱 +掂 +愚 +泷 +涪 +漳 +妩 +娉 +榄 +讷 +觅 +旧 +藤 +煮 +呛 +柳 +腓 +叭 +庵 +烷 +阡 +罂 +蜕 +擂 +猖 +咿 +媲 +脉 +【 +沏 +貅 +黠 +熏 +哲 +烁 +坦 +酵 +兜 +× +潇 +撒 +剽 +珩 +圹 +乾 +摸 +樟 +帽 +嗒 +襄 +魂 +轿 +憬 +锡 +〕 +喃 +皆 +咖 +隅 +脸 +残 +泮 +袂 +鹂 +珊 +囤 +捆 +咤 +误 +徨 +闹 +淙 +芊 +淋 +怆 +囗 +拨 +梳 +渤 +R +G +绨 +蚓 +婀 +幡 +狩 +麾 +谢 +唢 +裸 +旌 +伉 +纶 +裂 +驳 +砼 +咛 +澄 +樨 +蹈 +宙 +澍 +倍 +貔 +操 +勇 +蟠 +摈 +砧 +虬 +够 +缁 +悦 +藿 +撸 +艹 +摁 +淹 +豇 +虎 +榭 +ˉ +吱 +d +° +喧 +荀 +踱 +侮 +奋 +偕 +饷 +犍 +惮 +坑 +璎 +徘 +宛 +妆 +袈 +倩 +窦 +昂 +荏 +乖 +K +怅 +撰 +鳙 +牙 +袁 +酞 +X +痿 +琼 +闸 +雁 +趾 +荚 +虻 +涝 +《 +杏 +韭 +偈 +烤 +绫 +鞘 +卉 +症 +遢 +蓥 +诋 +杭 +荨 +匆 +竣 +簪 +辙 +敕 +虞 +丹 +缭 +咩 +黟 +m +淤 +瑕 +咂 +铉 +硼 +茨 +嶂 +痒 +畸 +敬 +涿 +粪 +窘 +熟 +叔 +嫔 +盾 +忱 +裘 +憾 +梵 +赡 +珙 +咯 +娘 +庙 +溯 +胺 +葱 +痪 +摊 +荷 +卞 +乒 +髦 +寐 +铭 +坩 +胗 +枷 +爆 +溟 +嚼 +羚 +砬 +轨 +惊 +挠 +罄 +竽 +菏 +氧 +浅 +楣 +盼 +枢 +炸 +阆 +杯 +谏 +噬 +淇 +渺 +俪 +秆 +墓 +泪 +跻 +砌 +痰 +垡 +渡 +耽 +釜 +讶 +鳎 +煞 +呗 +韶 +舶 +绷 +鹳 +缜 +旷 +铊 +皱 +龌 +檀 +霖 +奄 +槐 +艳 +蝶 +旋 +哝 +赶 +骞 +蚧 +腊 +盈 +丁 +` +蜚 +矸 +蝙 +睨 +嚓 +僻 +鬼 +醴 +夜 +彝 +磊 +笔 +拔 +栀 +糕 +厦 +邰 +纫 +逭 +纤 +眦 +膊 +馍 +躇 +烯 +蘼 +冬 +诤 +暄 +骶 +哑 +瘠 +」 +臊 +丕 +愈 +咱 +螺 +擅 +跋 +搏 +硪 +谄 +笠 +淡 +嘿 +骅 +谧 +鼎 +皋 +姚 +歼 +蠢 +驼 +耳 +胬 +挝 +涯 +狗 +蒽 +孓 +犷 +凉 +芦 +箴 +铤 +孤 +嘛 +坤 +V +茴 +朦 +挞 +尖 +橙 +诞 +搴 +碇 +洵 +浚 +帚 +蜍 +漯 +柘 +嚎 +讽 +芭 +荤 +咻 +祠 +秉 +跖 +埃 +吓 +糯 +眷 +馒 +惹 +娼 +鲑 +嫩 +讴 +轮 +瞥 +靶 +褚 +乏 +缤 +宋 +帧 +删 +驱 +碎 +扑 +俩 +俄 +偏 +涣 +竹 +噱 +皙 +佰 +渚 +唧 +斡 +# +镉 +刀 +崎 +筐 +佣 +夭 +贰 +肴 +峙 +哔 +艿 +匐 +牺 +镛 +缘 +仡 +嫡 +劣 +枸 +堀 +梨 +簿 +鸭 +蒸 +亦 +稽 +浴 +{ +衢 +束 +槲 +j +阁 +揍 +疥 +棋 +潋 +聪 +窜 +乓 +睛 +插 +冉 +阪 +苍 +搽 +「 +蟾 +螟 +幸 +仇 +樽 +撂 +慢 +跤 +幔 +俚 +淅 +覃 +觊 +溶 +妖 +帛 +侨 +曰 +妾 +泗 +· +： +瀘 +風 +Ë +（ +） +∶ +紅 +紗 +瑭 +雲 +頭 +鶏 +財 +許 +• +¥ +樂 +焗 +麗 +— +； +滙 +東 +榮 +繪 +興 +… +門 +業 +π +楊 +國 +顧 +é +盤 +寳 +Λ +龍 +鳳 +島 +誌 +緣 +結 +銭 +萬 +勝 +祎 +璟 +優 +歡 +臨 +時 +購 +＝ +★ +藍 +昇 +鐵 +觀 +勅 +農 +聲 +畫 +兿 +術 +發 +劉 +記 +專 +耑 +園 +書 +壴 +種 +Ο +● +褀 +號 +銀 +匯 +敟 +锘 +葉 +橪 +廣 +進 +蒄 +鑽 +阝 +祙 +貢 +鍋 +豊 +夬 +喆 +團 +閣 +開 +燁 +賓 +館 +酡 +沔 +順 +＋ +硚 +劵 +饸 +陽 +車 +湓 +復 +萊 +氣 +軒 +華 +堃 +迮 +纟 +戶 +馬 +學 +裡 +電 +嶽 +獨 +マ +シ +サ +ジ +燘 +袪 +環 +❤ +臺 +灣 +専 +賣 +孖 +聖 +攝 +線 +▪ +α +傢 +俬 +夢 +達 +莊 +喬 +貝 +薩 +劍 +羅 +壓 +棛 +饦 +尃 +璈 +囍 +醫 +Ｇ +Ｉ +Ａ +＃ +Ｎ +鷄 +髙 +嬰 +啓 +約 +隹 +潔 +賴 +藝 +～ +寶 +籣 +麺 +　 +嶺 +√ +義 +網 +峩 +長 +∧ +魚 +機 +構 +② +鳯 +偉 +Ｌ +Ｂ +㙟 +畵 +鴿 +＇ +詩 +溝 +嚞 +屌 +藔 +佧 +玥 +蘭 +織 +１ +３ +９ +０ +７ +點 +砭 +鴨 +鋪 +銘 +廳 +弍 +‧ +創 +湯 +坶 +℃ +卩 +骝 +＆ +烜 +荘 +當 +潤 +扞 +係 +懷 +碶 +钅 +蚨 +讠 +☆ +叢 +爲 +埗 +涫 +塗 +→ +楽 +現 +鯨 +愛 +瑪 +鈺 +忄 +悶 +藥 +飾 +樓 +視 +孬 +ㆍ +燚 +苪 +師 +① +丼 +锽 +│ +韓 +標 +è +兒 +閏 +匋 +張 +漢 +Ü +髪 +會 +閑 +檔 +習 +裝 +の +峯 +菘 +輝 +И +雞 +釣 +億 +浐 +Ｋ +Ｏ +Ｒ +８ +Ｈ +Ｅ +Ｐ +Ｔ +Ｗ +Ｄ +Ｓ +Ｃ +Ｍ +Ｆ +姌 +饹 +» +晞 +廰 +ä +嵯 +鷹 +負 +飲 +絲 +冚 +楗 +澤 +綫 +區 +❋ +← +質 +靑 +揚 +③ +滬 +統 +産 +協 +﹑ +乸 +畐 +經 +運 +際 +洺 +岽 +為 +粵 +諾 +崋 +豐 +碁 +ɔ +Ｖ +２ +６ +齋 +誠 +訂 +´ +勑 +雙 +陳 +無 +í +泩 +媄 +夌 +刂 +ｉ +ｃ +ｔ +ｏ +ｒ +ａ +嘢 +耄 +燴 +暃 +壽 +媽 +靈 +抻 +體 +唻 +É +冮 +甹 +鎮 +錦 +ʌ +蜛 +蠄 +尓 +駕 +戀 +飬 +逹 +倫 +貴 +極 +Я +Й +寬 +磚 +嶪 +郎 +職 +｜ +間 +ｎ +ｄ +剎 +伈 +課 +飛 +橋 +瘊 +№ +譜 +骓 +圗 +滘 +縣 +粿 +咅 +養 +濤 +彳 +® +％ +Ⅱ +啰 +㴪 +見 +矞 +薬 +糁 +邨 +鲮 +顔 +罱 +З +選 +話 +贏 +氪 +俵 +競 +瑩 +繡 +枱 +β +綉 +á +獅 +爾 +™ +麵 +戋 +淩 +徳 +個 +劇 +場 +務 +簡 +寵 +ｈ +實 +膠 +轱 +圖 +築 +嘣 +樹 +㸃 +營 +耵 +孫 +饃 +鄺 +飯 +麯 +遠 +輸 +坫 +孃 +乚 +閃 +鏢 +㎡ +題 +廠 +關 +↑ +爺 +將 +軍 +連 +篦 +覌 +參 +箸 +－ +窠 +棽 +寕 +夀 +爰 +歐 +呙 +閥 +頡 +熱 +雎 +垟 +裟 +凬 +勁 +帑 +馕 +夆 +疌 +枼 +馮 +貨 +蒤 +樸 +彧 +旸 +靜 +龢 +暢 +㐱 +鳥 +珺 +鏡 +灡 +爭 +堷 +廚 +Ó +騰 +診 +┅ +蘇 +褔 +凱 +頂 +豕 +亞 +帥 +嘬 +⊥ +仺 +桖 +複 +饣 +絡 +穂 +顏 +棟 +納 +▏ +濟 +親 +設 +計 +攵 +埌 +烺 +ò +頤 +燦 +蓮 +撻 +節 +講 +濱 +濃 +娽 +洳 +朿 +燈 +鈴 +護 +膚 +铔 +過 +補 +Ｚ +Ｕ +５ +４ +坋 +闿 +䖝 +餘 +缐 +铞 +貿 +铪 +桼 +趙 +鍊 +［ +㐂 +垚 +菓 +揸 +捲 +鐘 +滏 +𣇉 +爍 +輪 +燜 +鴻 +鮮 +動 +鹞 +鷗 +丄 +慶 +鉌 +翥 +飮 +腸 +⇋ +漁 +覺 +來 +熘 +昴 +翏 +鲱 +圧 +鄉 +萭 +頔 +爐 +嫚 +г +貭 +類 +聯 +幛 +輕 +訓 +鑒 +夋 +锨 +芃 +珣 +䝉 +扙 +嵐 +銷 +處 +ㄱ +語 +誘 +苝 +歸 +儀 +燒 +楿 +內 +粢 +葒 +奧 +麥 +礻 +滿 +蠔 +穵 +瞭 +態 +鱬 +榞 +硂 +鄭 +黃 +煙 +祐 +奓 +逺 +＊ +瑄 +獲 +聞 +薦 +讀 +這 +樣 +決 +問 +啟 +們 +執 +説 +轉 +單 +隨 +唘 +帶 +倉 +庫 +還 +贈 +尙 +皺 +■ +餅 +產 +○ +∈ +報 +狀 +楓 +賠 +琯 +嗮 +禮 +｀ +傳 +＞ +≤ +嗞 +Φ +≥ +換 +咭 +∣ +↓ +曬 +ε +応 +寫 +″ +終 +様 +純 +費 +療 +聨 +凍 +壐 +郵 +ü +黒 +∫ +製 +塊 +調 +軽 +確 +撃 +級 +馴 +Ⅲ +涇 +繹 +數 +碼 +證 +狒 +処 +劑 +＜ +晧 +賀 +衆 +］ +櫥 +兩 +陰 +絶 +對 +鯉 +憶 +◎ +ｐ +ｅ +Ｙ +蕒 +煖 +頓 +測 +試 +鼽 +僑 +碩 +妝 +帯 +≈ +鐡 +舖 +權 +喫 +倆 +ˋ +該 +悅 +ā +俫 +． +ｆ +ｓ +ｂ +ｍ +ｋ +ｇ +ｕ +ｊ +貼 +淨 +濕 +針 +適 +備 +ｌ +／ +給 +謢 +強 +觸 +衛 +與 +⊙ +＄ +緯 +變 +⑴ +⑵ +⑶ +㎏ +殺 +∩ +幚 +─ +價 +▲ +離 +ú +ó +飄 +烏 +関 +閟 +﹝ +﹞ +邏 +輯 +鍵 +驗 +訣 +導 +歷 +屆 +層 +▼ +儱 +錄 +熳 +ē +艦 +吋 +錶 +辧 +飼 +顯 +④ +禦 +販 +気 +対 +枰 +閩 +紀 +幹 +瞓 +貊 +淚 +△ +眞 +墊 +Ω +獻 +褲 +縫 +緑 +亜 +鉅 +餠 +｛ +｝ +◆ +蘆 +薈 +█ +◇ +溫 +彈 +晳 +粧 +犸 +穩 +訊 +崬 +凖 +熥 +П +舊 +條 +紋 +圍 +Ⅳ +筆 +尷 +難 +雜 +錯 +綁 +識 +頰 +鎖 +艶 +□ +殁 +殼 +⑧ +├ +▕ +鵬 +ǐ +ō +ǒ +糝 +綱 +▎ +μ +盜 +饅 +醬 +籤 +蓋 +釀 +鹽 +據 +à +ɡ +辦 +◥ +彐 +┌ +婦 +獸 +鲩 +伱 +ī +蒟 +蒻 +齊 +袆 +腦 +寧 +凈 +妳 +煥 +詢 +偽 +謹 +啫 +鯽 +騷 +鱸 +損 +傷 +鎻 +髮 +買 +冏 +儥 +両 +﹢ +∞ +載 +喰 +ｚ +羙 +悵 +燙 +曉 +員 +組 +徹 +艷 +痠 +鋼 +鼙 +縮 +細 +嚒 +爯 +≠ +維 +＂ +鱻 +壇 +厍 +帰 +浥 +犇 +薡 +軎 +² +應 +醜 +刪 +緻 +鶴 +賜 +噁 +軌 +尨 +镔 +鷺 +槗 +彌 +葚 +濛 +請 +溇 +緹 +賢 +訪 +獴 +瑅 +資 +縤 +陣 +蕟 +栢 +韻 +祼 +恁 +伢 +謝 +劃 +涑 +總 +衖 +踺 +砋 +凉 +籃 +駿 +苼 +瘋 +昽 +紡 +驊 +腎 +﹗ +響 +杋 +剛 +嚴 +禪 +歓 +槍 +傘 +檸 +檫 +炣 +勢 +鏜 +鎢 +銑 +尐 +減 +奪 +惡 +θ +僮 +婭 +臘 +ū +ì +殻 +鉄 +∑ +蛲 +焼 +緖 +續 +紹 +懮 \ No newline at end of file diff --git a/pp_onnx/onnx_paddleocr.py b/pp_onnx/onnx_paddleocr.py new file mode 100644 index 0000000..b7b6f67 --- /dev/null +++ b/pp_onnx/onnx_paddleocr.py @@ -0,0 +1,94 @@ +import time + +from pp_onnx.predict_system import TextSystem +from pp_onnx.utils import infer_args as init_args +from pp_onnx.utils import str2bool, draw_ocr +import argparse +import sys + + +class ONNXPaddleOcr(TextSystem): + def __init__(self, **kwargs): + # 默认参数 + parser = init_args() + # import IPython + # IPython.embed(header='L-14') + + inference_args_dict = {} + for action in parser._actions: + inference_args_dict[action.dest] = action.default + params = argparse.Namespace(**inference_args_dict) + + params.rec_image_shape = "3, 48, 320" + + # 根据传入的参数覆盖更新默认参数 + params.__dict__.update(**kwargs) + + # 初始化模型 + super().__init__(params) + + def ocr(self, img, det=True, rec=True, cls=True): + if cls == True and self.use_angle_cls == False: + print('Since the angle classifier is not initialized, the angle classifier will not be uesd during the forward process') + + if det and rec: + ocr_res = [] + dt_boxes, rec_res = self.__call__(img, cls) + tmp_res = [[box.tolist(), res] for box, res in zip(dt_boxes, rec_res)] + ocr_res.append(tmp_res) + return ocr_res + elif det and not rec: + ocr_res = [] + dt_boxes = self.text_detector(img) + tmp_res = [box.tolist() for box in dt_boxes] + ocr_res.append(tmp_res) + return ocr_res + else: + ocr_res = [] + cls_res = [] + + if not isinstance(img, list): + img = [img] + if self.use_angle_cls and cls: + img, cls_res_tmp = self.text_classifier(img) + if not rec: + cls_res.append(cls_res_tmp) + rec_res = self.text_recognizer(img) + ocr_res.append(rec_res) + + if not rec: + return cls_res + return ocr_res + + +def sav2Img(org_img, result, name="draw_ocr.jpg"): + # 显示结果 + from PIL import Image + result = result[0] + # image = Image.open(img_path).convert('RGB') + # 图像转BGR2RGB + image = org_img[:, :, ::-1] + boxes = [line[0] for line in result] + txts = [line[1][0] for line in result] + scores = [line[1][1] for line in result] + im_show = draw_ocr(image, boxes, txts, scores) + im_show = Image.fromarray(im_show) + im_show.save(name) + + +if __name__ == '__main__': + import cv2 + + model = ONNXPaddleOcr(use_angle_cls=True, use_gpu=False) + + + img = cv2.imread('/data2/liujingsong3/fiber_box/test/img/20230531230052008263304.jpg') + s = time.time() + result = model.ocr(img) + e = time.time() + print("total time: {:.3f}".format(e - s)) + print("result:", result) + for box in result[0]: + print(box) + + sav2Img(img, result) \ No newline at end of file diff --git a/pp_onnx/operators.py b/pp_onnx/operators.py new file mode 100644 index 0000000..86e122d --- /dev/null +++ b/pp_onnx/operators.py @@ -0,0 +1,187 @@ +import numpy as np +import cv2 +import sys +import math + + +class NormalizeImage(object): + """ normalize image such as substract mean, divide std + """ + + def __init__(self, scale=None, mean=None, std=None, order='chw', **kwargs): + if isinstance(scale, str): + scale = eval(scale) + self.scale = np.float32(scale if scale is not None else 1.0 / 255.0) + mean = mean if mean is not None else [0.485, 0.456, 0.406] + std = std if std is not None else [0.229, 0.224, 0.225] + + shape = (3, 1, 1) if order == 'chw' else (1, 1, 3) + self.mean = np.array(mean).reshape(shape).astype('float32') + self.std = np.array(std).reshape(shape).astype('float32') + + def __call__(self, data): + img = data['image'] + from PIL import Image + if isinstance(img, Image.Image): + img = np.array(img) + assert isinstance(img, + np.ndarray), "invalid input 'img' in NormalizeImage" + data['image'] = ( + img.astype('float32') * self.scale - self.mean) / self.std + return data + + +class DetResizeForTest(object): + def __init__(self, **kwargs): + super(DetResizeForTest, self).__init__() + self.resize_type = 0 + self.keep_ratio = False + if 'image_shape' in kwargs: + self.image_shape = kwargs['image_shape'] + self.resize_type = 1 + if 'keep_ratio' in kwargs: + self.keep_ratio = kwargs['keep_ratio'] + elif 'limit_side_len' in kwargs: + self.limit_side_len = kwargs['limit_side_len'] + self.limit_type = kwargs.get('limit_type', 'min') + elif 'resize_long' in kwargs: + self.resize_type = 2 + self.resize_long = kwargs.get('resize_long', 960) + else: + self.limit_side_len = 736 + self.limit_type = 'min' + + def __call__(self, data): + img = data['image'] + src_h, src_w, _ = img.shape + if sum([src_h, src_w]) < 64: + img = self.image_padding(img) + + if self.resize_type == 0: + # img, shape = self.resize_image_type0(img) + img, [ratio_h, ratio_w] = self.resize_image_type0(img) + elif self.resize_type == 2: + img, [ratio_h, ratio_w] = self.resize_image_type2(img) + else: + # img, shape = self.resize_image_type1(img) + img, [ratio_h, ratio_w] = self.resize_image_type1(img) + data['image'] = img + data['shape'] = np.array([src_h, src_w, ratio_h, ratio_w]) + return data + + def image_padding(self, im, value=0): + h, w, c = im.shape + im_pad = np.zeros((max(32, h), max(32, w), c), np.uint8) + value + im_pad[:h, :w, :] = im + return im_pad + + def resize_image_type1(self, img): + resize_h, resize_w = self.image_shape + ori_h, ori_w = img.shape[:2] # (h, w, c) + if self.keep_ratio is True: + resize_w = ori_w * resize_h / ori_h + N = math.ceil(resize_w / 32) + resize_w = N * 32 + ratio_h = float(resize_h) / ori_h + ratio_w = float(resize_w) / ori_w + img = cv2.resize(img, (int(resize_w), int(resize_h))) + # return img, np.array([ori_h, ori_w]) + return img, [ratio_h, ratio_w] + + def resize_image_type0(self, img): + """ + resize image to a size multiple of 32 which is required by the network + args: + img(array): array with shape [h, w, c] + return(tuple): + img, (ratio_h, ratio_w) + """ + limit_side_len = self.limit_side_len + h, w, c = img.shape + + # limit the max side + if self.limit_type == 'max': + if max(h, w) > limit_side_len: + if h > w: + ratio = float(limit_side_len) / h + else: + ratio = float(limit_side_len) / w + else: + ratio = 1. + elif self.limit_type == 'min': + if min(h, w) < limit_side_len: + if h < w: + ratio = float(limit_side_len) / h + else: + ratio = float(limit_side_len) / w + else: + ratio = 1. + elif self.limit_type == 'resize_long': + ratio = float(limit_side_len) / max(h, w) + else: + raise Exception('not support limit type, image ') + resize_h = int(h * ratio) + resize_w = int(w * ratio) + + resize_h = max(int(round(resize_h / 32) * 32), 32) + resize_w = max(int(round(resize_w / 32) * 32), 32) + + try: + if int(resize_w) <= 0 or int(resize_h) <= 0: + return None, (None, None) + img = cv2.resize(img, (int(resize_w), int(resize_h))) + except: + print(img.shape, resize_w, resize_h) + sys.exit(0) + ratio_h = resize_h / float(h) + ratio_w = resize_w / float(w) + return img, [ratio_h, ratio_w] + + def resize_image_type2(self, img): + h, w, _ = img.shape + + resize_w = w + resize_h = h + + if resize_h > resize_w: + ratio = float(self.resize_long) / resize_h + else: + ratio = float(self.resize_long) / resize_w + + resize_h = int(resize_h * ratio) + resize_w = int(resize_w * ratio) + + max_stride = 128 + resize_h = (resize_h + max_stride - 1) // max_stride * max_stride + resize_w = (resize_w + max_stride - 1) // max_stride * max_stride + img = cv2.resize(img, (int(resize_w), int(resize_h))) + ratio_h = resize_h / float(h) + ratio_w = resize_w / float(w) + + return img, [ratio_h, ratio_w] + +class ToCHWImage(object): + """ convert hwc image to chw image + """ + + def __init__(self, **kwargs): + pass + + def __call__(self, data): + img = data['image'] + from PIL import Image + if isinstance(img, Image.Image): + img = np.array(img) + data['image'] = img.transpose((2, 0, 1)) + return data + + +class KeepKeys(object): + def __init__(self, keep_keys, **kwargs): + self.keep_keys = keep_keys + + def __call__(self, data): + data_list = [] + for key in self.keep_keys: + data_list.append(data[key]) + return data_list \ No newline at end of file diff --git a/pp_onnx/predict_base.py b/pp_onnx/predict_base.py new file mode 100644 index 0000000..c3eef36 --- /dev/null +++ b/pp_onnx/predict_base.py @@ -0,0 +1,52 @@ +import onnxruntime + +class PredictBase(object): + def __init__(self): + pass + + def get_onnx_session(self, model_dir, use_gpu): + # 使用gpu + if use_gpu: + providers = providers=['CUDAExecutionProvider'] + else: + providers = providers = ['CPUExecutionProvider'] + + onnx_session = onnxruntime.InferenceSession(str(model_dir), None, providers=providers) + + # print("providers:", onnxruntime.get_device()) + return onnx_session + + + def get_output_name(self, onnx_session): + """ + output_name = onnx_session.get_outputs()[0].name + :param onnx_session: + :return: + """ + output_name = [] + for node in onnx_session.get_outputs(): + output_name.append(node.name) + return output_name + + def get_input_name(self, onnx_session): + """ + input_name = onnx_session.get_inputs()[0].name + :param onnx_session: + :return: + """ + input_name = [] + for node in onnx_session.get_inputs(): + input_name.append(node.name) + return input_name + + def get_input_feed(self, input_name, image_numpy): + """ + input_feed={self.input_name: image_numpy} + :param input_name: + :param image_numpy: + :return: + """ + input_feed = {} + for name in input_name: + input_feed[name] = image_numpy + return input_feed \ No newline at end of file diff --git a/pp_onnx/predict_cls.py b/pp_onnx/predict_cls.py new file mode 100644 index 0000000..c9e5cb1 --- /dev/null +++ b/pp_onnx/predict_cls.py @@ -0,0 +1,86 @@ +import cv2 +import copy +import numpy as np +import math + +from pp_onnx.cls_postprocess import ClsPostProcess +from pp_onnx.predict_base import PredictBase + +class TextClassifier(PredictBase): + def __init__(self, args): + self.cls_image_shape = [int(v) for v in args.cls_image_shape.split(",")] + self.cls_batch_num = args.cls_batch_num + self.cls_thresh = args.cls_thresh + self.postprocess_op = ClsPostProcess(label_list=args.label_list) + + # 初始化模型 + self.cls_onnx_session = self.get_onnx_session(args.cls_model_dir, args.use_gpu) + self.cls_input_name = self.get_input_name(self.cls_onnx_session) + self.cls_output_name = self.get_output_name(self.cls_onnx_session) + + def resize_norm_img(self, img): + imgC, imgH, imgW = self.cls_image_shape + h = img.shape[0] + w = img.shape[1] + ratio = w / float(h) + if math.ceil(imgH * ratio) > imgW: + resized_w = imgW + else: + resized_w = int(math.ceil(imgH * ratio)) + resized_image = cv2.resize(img, (resized_w, imgH)) + resized_image = resized_image.astype('float32') + if self.cls_image_shape[0] == 1: + resized_image = resized_image / 255 + resized_image = resized_image[np.newaxis, :] + else: + resized_image = resized_image.transpose((2, 0, 1)) / 255 + resized_image -= 0.5 + resized_image /= 0.5 + padding_im = np.zeros((imgC, imgH, imgW), dtype=np.float32) + padding_im[:, :, 0:resized_w] = resized_image + return padding_im + + def __call__(self, img_list): + img_list = copy.deepcopy(img_list) + img_num = len(img_list) + # Calculate the aspect ratio of all text bars + width_list = [] + for img in img_list: + width_list.append(img.shape[1] / float(img.shape[0])) + # Sorting can speed up the cls process + indices = np.argsort(np.array(width_list)) + + cls_res = [['', 0.0]] * img_num + batch_num = self.cls_batch_num + + for beg_img_no in range(0, img_num, batch_num): + + end_img_no = min(img_num, beg_img_no + batch_num) + norm_img_batch = [] + max_wh_ratio = 0 + + for ino in range(beg_img_no, end_img_no): + h, w = img_list[indices[ino]].shape[0:2] + wh_ratio = w * 1.0 / h + max_wh_ratio = max(max_wh_ratio, wh_ratio) + for ino in range(beg_img_no, end_img_no): + norm_img = self.resize_norm_img(img_list[indices[ino]]) + norm_img = norm_img[np.newaxis, :] + norm_img_batch.append(norm_img) + norm_img_batch = np.concatenate(norm_img_batch) + norm_img_batch = norm_img_batch.copy() + + input_feed = self.get_input_feed(self.cls_input_name, norm_img_batch) + outputs = self.cls_onnx_session.run(self.cls_output_name, input_feed=input_feed) + + prob_out = outputs[0] + + cls_result = self.postprocess_op(prob_out) + for rno in range(len(cls_result)): + label, score = cls_result[rno] + cls_res[indices[beg_img_no + rno]] = [label, score] + if '180' in label and score > self.cls_thresh: + img_list[indices[beg_img_no + rno]] = cv2.rotate( + img_list[indices[beg_img_no + rno]], 1) + return img_list, cls_res + diff --git a/pp_onnx/predict_det.py b/pp_onnx/predict_det.py new file mode 100644 index 0000000..2f6fc88 --- /dev/null +++ b/pp_onnx/predict_det.py @@ -0,0 +1,126 @@ +import numpy as np +from pp_onnx.imaug import transform, create_operators +from pp_onnx.db_postprocess import DBPostProcess +from pp_onnx.predict_base import PredictBase + + +class TextDetector(PredictBase): + def __init__(self, args): + self.args = args + self.det_algorithm = args.det_algorithm + pre_process_list = [{ + 'DetResizeForTest': { + 'limit_side_len': args.det_limit_side_len, + 'limit_type': args.det_limit_type, + } + }, { + 'NormalizeImage': { + 'std': [0.229, 0.224, 0.225], + 'mean': [0.485, 0.456, 0.406], + 'scale': '1./255.', + 'order': 'hwc' + } + }, { + 'ToCHWImage': None + }, { + 'KeepKeys': { + 'keep_keys': ['image', 'shape'] + } + }] + postprocess_params = {} + postprocess_params['name'] = 'DBPostProcess' + postprocess_params["thresh"] = args.det_db_thresh + postprocess_params["box_thresh"] = args.det_db_box_thresh + postprocess_params["max_candidates"] = 1000 + postprocess_params["unclip_ratio"] = args.det_db_unclip_ratio + postprocess_params["use_dilation"] = args.use_dilation + postprocess_params["score_mode"] = args.det_db_score_mode + postprocess_params["box_type"] = args.det_box_type + + # 实例化预处理操作类 + self.preprocess_op = create_operators(pre_process_list) + # self.postprocess_op = build_post_process(postprocess_params) + # 实例化后处理操作类 + self.postprocess_op = DBPostProcess(**postprocess_params) + + # 初始化模型 + self.det_onnx_session = self.get_onnx_session(args.det_model_dir, args.use_gpu) + self.det_input_name = self.get_input_name(self.det_onnx_session) + self.det_output_name = self.get_output_name(self.det_onnx_session) + + + + + def order_points_clockwise(self, pts): + rect = np.zeros((4, 2), dtype="float32") + s = pts.sum(axis=1) + rect[0] = pts[np.argmin(s)] + rect[2] = pts[np.argmax(s)] + tmp = np.delete(pts, (np.argmin(s), np.argmax(s)), axis=0) + diff = np.diff(np.array(tmp), axis=1) + rect[1] = tmp[np.argmin(diff)] + rect[3] = tmp[np.argmax(diff)] + return rect + + def clip_det_res(self, points, img_height, img_width): + for pno in range(points.shape[0]): + points[pno, 0] = int(min(max(points[pno, 0], 0), img_width - 1)) + points[pno, 1] = int(min(max(points[pno, 1], 0), img_height - 1)) + return points + + def filter_tag_det_res(self, dt_boxes, image_shape): + img_height, img_width = image_shape[0:2] + dt_boxes_new = [] + for box in dt_boxes: + if type(box) is list: + box = np.array(box) + box = self.order_points_clockwise(box) + box = self.clip_det_res(box, img_height, img_width) + rect_width = int(np.linalg.norm(box[0] - box[1])) + rect_height = int(np.linalg.norm(box[0] - box[3])) + if rect_width <= 3 or rect_height <= 3: + continue + dt_boxes_new.append(box) + dt_boxes = np.array(dt_boxes_new) + return dt_boxes + + def filter_tag_det_res_only_clip(self, dt_boxes, image_shape): + img_height, img_width = image_shape[0:2] + dt_boxes_new = [] + for box in dt_boxes: + if type(box) is list: + box = np.array(box) + box = self.clip_det_res(box, img_height, img_width) + dt_boxes_new.append(box) + dt_boxes = np.array(dt_boxes_new) + return dt_boxes + + def __call__(self, img): + ori_im = img.copy() + data = {'image': img} + + data = transform(data, self.preprocess_op) + img, shape_list = data + if img is None: + return None, 0 + img = np.expand_dims(img, axis=0) + shape_list = np.expand_dims(shape_list, axis=0) + img = img.copy() + + + input_feed = self.get_input_feed(self.det_input_name, img) + outputs = self.det_onnx_session.run(self.det_output_name, input_feed=input_feed) + + preds = {} + preds['maps'] = outputs[0] + + post_result = self.postprocess_op(preds, shape_list) + dt_boxes = post_result[0]['points'] + + if self.args.det_box_type == 'poly': + dt_boxes = self.filter_tag_det_res_only_clip(dt_boxes, ori_im.shape) + else: + dt_boxes = self.filter_tag_det_res(dt_boxes, ori_im.shape) + + return dt_boxes + diff --git a/pp_onnx/predict_rec.py b/pp_onnx/predict_rec.py new file mode 100644 index 0000000..5ddf7f6 --- /dev/null +++ b/pp_onnx/predict_rec.py @@ -0,0 +1,321 @@ +import cv2 +import numpy as np +import math +from PIL import Image + + +from pp_onnx.rec_postprocess import CTCLabelDecode +from pp_onnx.predict_base import PredictBase + +class TextRecognizer(PredictBase): + def __init__(self, args): + self.rec_image_shape = [int(v) for v in args.rec_image_shape.split(",")] + self.rec_batch_num = args.rec_batch_num + self.rec_algorithm = args.rec_algorithm + self.postprocess_op = CTCLabelDecode(character_dict_path=args.rec_char_dict_path, use_space_char=args.use_space_char) + + # 初始化模型 + self.rec_onnx_session = self.get_onnx_session(args.rec_model_dir, args.use_gpu) + self.rec_input_name = self.get_input_name(self.rec_onnx_session) + self.rec_output_name = self.get_output_name(self.rec_onnx_session) + + + def resize_norm_img(self, img, max_wh_ratio): + imgC, imgH, imgW = self.rec_image_shape + if self.rec_algorithm == 'NRTR' or self.rec_algorithm == 'ViTSTR': + img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) + # return padding_im + image_pil = Image.fromarray(np.uint8(img)) + if self.rec_algorithm == 'ViTSTR': + img = image_pil.resize([imgW, imgH], Image.BICUBIC) + else: + img = image_pil.resize([imgW, imgH], Image.ANTIALIAS) + img = np.array(img) + norm_img = np.expand_dims(img, -1) + norm_img = norm_img.transpose((2, 0, 1)) + if self.rec_algorithm == 'ViTSTR': + norm_img = norm_img.astype(np.float32) / 255. + else: + norm_img = norm_img.astype(np.float32) / 128. - 1. + return norm_img + elif self.rec_algorithm == 'RFL': + img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) + resized_image = cv2.resize( + img, (imgW, imgH), interpolation=cv2.INTER_CUBIC) + resized_image = resized_image.astype('float32') + resized_image = resized_image / 255 + resized_image = resized_image[np.newaxis, :] + resized_image -= 0.5 + resized_image /= 0.5 + return resized_image + + assert imgC == img.shape[2] + imgW = int((imgH * max_wh_ratio)) + + # import IPython + # IPython.embed(header="predict_rec.py L-56") + + w = self.rec_onnx_session.get_inputs()[0].shape[3:][0] + if isinstance(w, int) and w>0: + imgW = w + # if w is not None and w > 0: + # imgW = w + + + h, w = img.shape[:2] + ratio = w / float(h) + if math.ceil(imgH * ratio) > imgW: + resized_w = imgW + else: + resized_w = int(math.ceil(imgH * ratio)) + if self.rec_algorithm == 'RARE': + if resized_w > self.rec_image_shape[2]: + resized_w = self.rec_image_shape[2] + imgW = self.rec_image_shape[2] + resized_image = cv2.resize(img, (resized_w, imgH)) + resized_image = resized_image.astype('float32') + resized_image = resized_image.transpose((2, 0, 1)) / 255 + resized_image -= 0.5 + resized_image /= 0.5 + padding_im = np.zeros((imgC, imgH, imgW), dtype=np.float32) + padding_im[:, :, 0:resized_w] = resized_image + return padding_im + + def resize_norm_img_vl(self, img, image_shape): + + imgC, imgH, imgW = image_shape + img = img[:, :, ::-1] # bgr2rgb + resized_image = cv2.resize( + img, (imgW, imgH), interpolation=cv2.INTER_LINEAR) + resized_image = resized_image.astype('float32') + resized_image = resized_image.transpose((2, 0, 1)) / 255 + return resized_image + + def resize_norm_img_srn(self, img, image_shape): + imgC, imgH, imgW = image_shape + + img_black = np.zeros((imgH, imgW)) + im_hei = img.shape[0] + im_wid = img.shape[1] + + if im_wid <= im_hei * 1: + img_new = cv2.resize(img, (imgH * 1, imgH)) + elif im_wid <= im_hei * 2: + img_new = cv2.resize(img, (imgH * 2, imgH)) + elif im_wid <= im_hei * 3: + img_new = cv2.resize(img, (imgH * 3, imgH)) + else: + img_new = cv2.resize(img, (imgW, imgH)) + + img_np = np.asarray(img_new) + img_np = cv2.cvtColor(img_np, cv2.COLOR_BGR2GRAY) + img_black[:, 0:img_np.shape[1]] = img_np + img_black = img_black[:, :, np.newaxis] + + row, col, c = img_black.shape + c = 1 + + return np.reshape(img_black, (c, row, col)).astype(np.float32) + + def srn_other_inputs(self, image_shape, num_heads, max_text_length): + + imgC, imgH, imgW = image_shape + feature_dim = int((imgH / 8) * (imgW / 8)) + + encoder_word_pos = np.array(range(0, feature_dim)).reshape( + (feature_dim, 1)).astype('int64') + gsrm_word_pos = np.array(range(0, max_text_length)).reshape( + (max_text_length, 1)).astype('int64') + + gsrm_attn_bias_data = np.ones((1, max_text_length, max_text_length)) + gsrm_slf_attn_bias1 = np.triu(gsrm_attn_bias_data, 1).reshape( + [-1, 1, max_text_length, max_text_length]) + gsrm_slf_attn_bias1 = np.tile( + gsrm_slf_attn_bias1, + [1, num_heads, 1, 1]).astype('float32') * [-1e9] + + gsrm_slf_attn_bias2 = np.tril(gsrm_attn_bias_data, -1).reshape( + [-1, 1, max_text_length, max_text_length]) + gsrm_slf_attn_bias2 = np.tile( + gsrm_slf_attn_bias2, + [1, num_heads, 1, 1]).astype('float32') * [-1e9] + + encoder_word_pos = encoder_word_pos[np.newaxis, :] + gsrm_word_pos = gsrm_word_pos[np.newaxis, :] + + return [ + encoder_word_pos, gsrm_word_pos, gsrm_slf_attn_bias1, + gsrm_slf_attn_bias2 + ] + + def process_image_srn(self, img, image_shape, num_heads, max_text_length): + norm_img = self.resize_norm_img_srn(img, image_shape) + norm_img = norm_img[np.newaxis, :] + + [encoder_word_pos, gsrm_word_pos, gsrm_slf_attn_bias1, gsrm_slf_attn_bias2] = \ + self.srn_other_inputs(image_shape, num_heads, max_text_length) + + gsrm_slf_attn_bias1 = gsrm_slf_attn_bias1.astype(np.float32) + gsrm_slf_attn_bias2 = gsrm_slf_attn_bias2.astype(np.float32) + encoder_word_pos = encoder_word_pos.astype(np.int64) + gsrm_word_pos = gsrm_word_pos.astype(np.int64) + + return (norm_img, encoder_word_pos, gsrm_word_pos, gsrm_slf_attn_bias1, + gsrm_slf_attn_bias2) + + def resize_norm_img_sar(self, img, image_shape, + width_downsample_ratio=0.25): + imgC, imgH, imgW_min, imgW_max = image_shape + h = img.shape[0] + w = img.shape[1] + valid_ratio = 1.0 + # make sure new_width is an integral multiple of width_divisor. + width_divisor = int(1 / width_downsample_ratio) + # resize + ratio = w / float(h) + resize_w = math.ceil(imgH * ratio) + if resize_w % width_divisor != 0: + resize_w = round(resize_w / width_divisor) * width_divisor + if imgW_min is not None: + resize_w = max(imgW_min, resize_w) + if imgW_max is not None: + valid_ratio = min(1.0, 1.0 * resize_w / imgW_max) + resize_w = min(imgW_max, resize_w) + resized_image = cv2.resize(img, (resize_w, imgH)) + resized_image = resized_image.astype('float32') + # norm + if image_shape[0] == 1: + resized_image = resized_image / 255 + resized_image = resized_image[np.newaxis, :] + else: + resized_image = resized_image.transpose((2, 0, 1)) / 255 + resized_image -= 0.5 + resized_image /= 0.5 + resize_shape = resized_image.shape + padding_im = -1.0 * np.ones((imgC, imgH, imgW_max), dtype=np.float32) + padding_im[:, :, 0:resize_w] = resized_image + pad_shape = padding_im.shape + + return padding_im, resize_shape, pad_shape, valid_ratio + + def resize_norm_img_spin(self, img): + img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) + # return padding_im + img = cv2.resize(img, tuple([100, 32]), cv2.INTER_CUBIC) + img = np.array(img, np.float32) + img = np.expand_dims(img, -1) + img = img.transpose((2, 0, 1)) + mean = [127.5] + std = [127.5] + mean = np.array(mean, dtype=np.float32) + std = np.array(std, dtype=np.float32) + mean = np.float32(mean.reshape(1, -1)) + stdinv = 1 / np.float32(std.reshape(1, -1)) + img -= mean + img *= stdinv + return img + + def resize_norm_img_svtr(self, img, image_shape): + + imgC, imgH, imgW = image_shape + resized_image = cv2.resize( + img, (imgW, imgH), interpolation=cv2.INTER_LINEAR) + resized_image = resized_image.astype('float32') + resized_image = resized_image.transpose((2, 0, 1)) / 255 + resized_image -= 0.5 + resized_image /= 0.5 + return resized_image + + def resize_norm_img_abinet(self, img, image_shape): + + imgC, imgH, imgW = image_shape + + resized_image = cv2.resize( + img, (imgW, imgH), interpolation=cv2.INTER_LINEAR) + resized_image = resized_image.astype('float32') + resized_image = resized_image / 255. + + mean = np.array([0.485, 0.456, 0.406]) + std = np.array([0.229, 0.224, 0.225]) + resized_image = ( + resized_image - mean[None, None, ...]) / std[None, None, ...] + resized_image = resized_image.transpose((2, 0, 1)) + resized_image = resized_image.astype('float32') + + return resized_image + + def norm_img_can(self, img, image_shape): + + img = cv2.cvtColor( + img, cv2.COLOR_BGR2GRAY) # CAN only predict gray scale image + + if self.inverse: + img = 255 - img + + if self.rec_image_shape[0] == 1: + h, w = img.shape + _, imgH, imgW = self.rec_image_shape + if h < imgH or w < imgW: + padding_h = max(imgH - h, 0) + padding_w = max(imgW - w, 0) + img_padded = np.pad(img, ((0, padding_h), (0, padding_w)), + 'constant', + constant_values=(255)) + img = img_padded + + img = np.expand_dims(img, 0) / 255.0 # h,w,c -> c,h,w + img = img.astype('float32') + + return img + + def __call__(self, img_list): + img_num = len(img_list) + # Calculate the aspect ratio of all text bars + width_list = [] + for img in img_list: + width_list.append(img.shape[1] / float(img.shape[0])) + # Sorting can speed up the recognition process + indices = np.argsort(np.array(width_list)) + rec_res = [['', 0.0]] * img_num + batch_num = self.rec_batch_num + + for beg_img_no in range(0, img_num, batch_num): + end_img_no = min(img_num, beg_img_no + batch_num) + norm_img_batch = [] + imgC, imgH, imgW = self.rec_image_shape[:3] + max_wh_ratio = imgW / imgH + # max_wh_ratio = 0 + for ino in range(beg_img_no, end_img_no): + h, w = img_list[indices[ino]].shape[0:2] + wh_ratio = w * 1.0 / h + max_wh_ratio = max(max_wh_ratio, wh_ratio) + for ino in range(beg_img_no, end_img_no): + norm_img = self.resize_norm_img(img_list[indices[ino]], + max_wh_ratio) + norm_img = norm_img[np.newaxis, :] + norm_img_batch.append(norm_img) + + norm_img_batch = np.concatenate(norm_img_batch) + norm_img_batch = norm_img_batch.copy() + + # img = img[:, :, ::-1].transpose(2, 0, 1) + # img = img[:, :, ::-1] + # img = img.transpose(2, 0, 1) + # img = img.astype(np.float32) + # img = np.expand_dims(img, axis=0) + # print(img.shape) + + input_feed = self.get_input_feed(self.rec_input_name, norm_img_batch) + + # import IPython + # IPython.embed(header='L-303') + + outputs = self.rec_onnx_session.run(self.rec_output_name, input_feed=input_feed) + + preds = outputs[0] + + rec_result = self.postprocess_op(preds) + for rno in range(len(rec_result)): + rec_res[indices[beg_img_no + rno]] = rec_result[rno] + + return rec_res diff --git a/pp_onnx/predict_system.py b/pp_onnx/predict_system.py new file mode 100644 index 0000000..702b71e --- /dev/null +++ b/pp_onnx/predict_system.py @@ -0,0 +1,99 @@ +import os +import cv2 +import copy +import pp_onnx.predict_det as predict_det +import pp_onnx.predict_cls as predict_cls +import pp_onnx.predict_rec as predict_rec +from pp_onnx.utils import get_rotate_crop_image, get_minarea_rect_crop + + +class TextSystem(object): + def __init__(self, args): + self.text_detector = predict_det.TextDetector(args) + self.text_recognizer = predict_rec.TextRecognizer(args) + self.use_angle_cls = args.use_angle_cls + self.drop_score = args.drop_score + if self.use_angle_cls: + self.text_classifier = predict_cls.TextClassifier(args) + + self.args = args + self.crop_image_res_index = 0 + + + def draw_crop_rec_res(self, output_dir, img_crop_list, rec_res): + os.makedirs(output_dir, exist_ok=True) + bbox_num = len(img_crop_list) + for bno in range(bbox_num): + cv2.imwrite( + os.path.join(output_dir, + f"mg_crop_{bno+self.crop_image_res_index}.jpg"), + img_crop_list[bno]) + + self.crop_image_res_index += bbox_num + + def __call__(self, img, cls=True): + ori_im = img.copy() + # 文字检测 + dt_boxes = self.text_detector(img) + + if dt_boxes is None: + return None, None + + img_crop_list = [] + + dt_boxes = sorted_boxes(dt_boxes) + + # 图片裁剪 + for bno in range(len(dt_boxes)): + tmp_box = copy.deepcopy(dt_boxes[bno]) + if self.args.det_box_type == "quad": + img_crop = get_rotate_crop_image(ori_im, tmp_box) + else: + img_crop = get_minarea_rect_crop(ori_im, tmp_box) + img_crop_list.append(img_crop) + + # 方向分类 + if self.use_angle_cls and cls: + img_crop_list, angle_list = self.text_classifier(img_crop_list) + + # 图像识别 + rec_res = self.text_recognizer(img_crop_list) + + if self.args.save_crop_res: + self.draw_crop_rec_res(self.args.crop_res_save_dir, img_crop_list,rec_res) + filter_boxes, filter_rec_res = [], [] + for box, rec_result in zip(dt_boxes, rec_res): + text, score = rec_result + if score >= self.drop_score: + filter_boxes.append(box) + filter_rec_res.append(rec_result) + + # import IPython + # IPython.embed(header='L-70') + + return filter_boxes, filter_rec_res + + +def sorted_boxes(dt_boxes): + """ + Sort text boxes in order from top to bottom, left to right + args: + dt_boxes(array):detected text boxes with shape [4, 2] + return: + sorted boxes(array) with shape [4, 2] + """ + num_boxes = dt_boxes.shape[0] + sorted_boxes = sorted(dt_boxes, key=lambda x: (x[0][1], x[0][0])) + _boxes = list(sorted_boxes) + + for i in range(num_boxes - 1): + for j in range(i, -1, -1): + if abs(_boxes[j + 1][0][1] - _boxes[j][0][1]) < 10 and \ + (_boxes[j + 1][0][0] < _boxes[j][0][0]): + tmp = _boxes[j] + _boxes[j] = _boxes[j + 1] + _boxes[j + 1] = tmp + else: + break + return _boxes + diff --git a/pp_onnx/readme.md b/pp_onnx/readme.md new file mode 100644 index 0000000..ba1d851 --- /dev/null +++ b/pp_onnx/readme.md @@ -0,0 +1,65 @@ +# paddleocr模型转换成onnx模型后，利用ONNX模型进行推理 +## 1、安装paddle2onnx +```angular2html +pip install paddle2onnx +``` + +## 2、下载paddleocr模型文件 +```angular2html +!wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar +!wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_rec_infer.tar +!wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_infer.tar +``` +## 3、解压模型文件 +```angular2html +!tar -xvf /home/aistudio/onnx_pred/models/ch_ppocr_mobile_v2.0_cls_infer.tar +!tar -xvf /home/aistudio/onnx_pred/models/ch_ppocr_server_v2.0_det_infer.tar +!tar -xvf /home/aistudio/onnx_pred/models/ch_ppocr_server_v2.0_rec_infer.tar +``` + +## 4、将paddleocr模型转成onxx模型 +```angular2html +paddle2onnx --model_dir ./ch_ppocr_server_v2.0_rec_infer \ +--model_filename inference.pdmodel \ +--params_filename inference.pdiparams \ +--save_file ./ch_ppocr_server_v2.0_rec.onnx \ +--opset_version 11 \ +--enable_onnx_checker True + + +paddle2onnx --model_dir ./ch_ppocr_server_v2.0_det_infer \ +--model_filename inference.pdmodel \ +--params_filename inference.pdiparams \ +--save_file ./ch_ppocr_server_v2.0_det.onnx \ +--opset_version 11 \ +--enable_onnx_checker True + + +paddle2onnx --model_dir ./ch_ppocr_mobile_v2.0_cls_infer \ +--model_filename inference.pdmodel \ +--params_filename inference.pdiparams \ +--save_file ./ch_ppocr_mobile_v2.0_cls.onnx \ +--opset_version 11 \ +--enable_onnx_checker True +``` + +## 5、安装onnx +```angular2html +pip install onnx==1.14.0 +pip install onnxruntime-gpu==1.14.1 +``` + +## 6、模型推理 +```angular2html + import cv2 + model = ONNXPaddleOcr() + + img = cv2.imread('./1.jpg') + + # ocr识别结果 + result = model.ocr(img) + print(result) + + # 画box框 + sav2Img(img, result) +``` \ No newline at end of file diff --git a/pp_onnx/rec_postprocess.py b/pp_onnx/rec_postprocess.py new file mode 100644 index 0000000..9c16bfb --- /dev/null +++ b/pp_onnx/rec_postprocess.py @@ -0,0 +1,920 @@ + +import numpy as np +# import paddle +# from paddle.nn import functional as F +import re + + +class BaseRecLabelDecode(object): + """ Convert between text-label and text-index """ + + def __init__(self, character_dict_path=None, use_space_char=False): + self.beg_str = "sos" + self.end_str = "eos" + self.reverse = False + self.character_str = [] + + if character_dict_path is None: + self.character_str = "0123456789abcdefghijklmnopqrstuvwxyz" + dict_character = list(self.character_str) + else: + with open(character_dict_path, "rb") as fin: + lines = fin.readlines() + for line in lines: + line = line.decode('utf-8').strip("\n").strip("\r\n") + self.character_str.append(line) + if use_space_char: + self.character_str.append(" ") + dict_character = list(self.character_str) + # import IPython + # IPython.embed(header='L-19') + if 'arabic' in str(character_dict_path): + self.reverse = True + + dict_character = self.add_special_char(dict_character) + self.dict = {} + for i, char in enumerate(dict_character): + self.dict[char] = i + self.character = dict_character + + def pred_reverse(self, pred): + pred_re = [] + c_current = '' + for c in pred: + if not bool(re.search('[a-zA-Z0-9 :*./%+-]', c)): + if c_current != '': + pred_re.append(c_current) + pred_re.append(c) + c_current = '' + else: + c_current += c + if c_current != '': + pred_re.append(c_current) + + return ''.join(pred_re[::-1]) + + def add_special_char(self, dict_character): + return dict_character + + def decode(self, text_index, text_prob=None, is_remove_duplicate=False): + """ convert text-index into text-label. """ + result_list = [] + ignored_tokens = self.get_ignored_tokens() + batch_size = len(text_index) + for batch_idx in range(batch_size): + selection = np.ones(len(text_index[batch_idx]), dtype=bool) + if is_remove_duplicate: + selection[1:] = text_index[batch_idx][1:] != text_index[ + batch_idx][:-1] + for ignored_token in ignored_tokens: + selection &= text_index[batch_idx] != ignored_token + + char_list = [ + self.character[text_id] + for text_id in text_index[batch_idx][selection] + ] + if text_prob is not None: + conf_list = text_prob[batch_idx][selection] + else: + conf_list = [1] * len(selection) + if len(conf_list) == 0: + conf_list = [0] + + text = ''.join(char_list) + + if self.reverse: # for arabic rec + text = self.pred_reverse(text) + + result_list.append((text, np.mean(conf_list).tolist())) + return result_list + + def get_ignored_tokens(self): + return [0] # for ctc blank + + +class CTCLabelDecode(BaseRecLabelDecode): + """ Convert between text-label and text-index """ + + def __init__(self, character_dict_path=None, use_space_char=False, + **kwargs): + super(CTCLabelDecode, self).__init__(character_dict_path, + use_space_char) + + def __call__(self, preds, label=None, *args, **kwargs): + if isinstance(preds, tuple) or isinstance(preds, list): + preds = preds[-1] + # if isinstance(preds, paddle.Tensor): + # preds = preds.numpy() + preds_idx = preds.argmax(axis=2) + preds_prob = preds.max(axis=2) + text = self.decode(preds_idx, preds_prob, is_remove_duplicate=True) + if label is None: + return text + label = self.decode(label) + return text, label + + def add_special_char(self, dict_character): + dict_character = ['blank'] + dict_character + return dict_character + + +class DistillationCTCLabelDecode(CTCLabelDecode): + """ + Convert + Convert between text-label and text-index + """ + + def __init__(self, + character_dict_path=None, + use_space_char=False, + model_name=["student"], + key=None, + multi_head=False, + **kwargs): + super(DistillationCTCLabelDecode, self).__init__(character_dict_path, + use_space_char) + if not isinstance(model_name, list): + model_name = [model_name] + self.model_name = model_name + + self.key = key + self.multi_head = multi_head + + def __call__(self, preds, label=None, *args, **kwargs): + output = dict() + for name in self.model_name: + pred = preds[name] + if self.key is not None: + pred = pred[self.key] + if self.multi_head and isinstance(pred, dict): + pred = pred['ctc'] + output[name] = super().__call__(pred, label=label, *args, **kwargs) + return output + + +class AttnLabelDecode(BaseRecLabelDecode): + """ Convert between text-label and text-index """ + + def __init__(self, character_dict_path=None, use_space_char=False, + **kwargs): + super(AttnLabelDecode, self).__init__(character_dict_path, + use_space_char) + + def add_special_char(self, dict_character): + self.beg_str = "sos" + self.end_str = "eos" + dict_character = dict_character + dict_character = [self.beg_str] + dict_character + [self.end_str] + return dict_character + + def decode(self, text_index, text_prob=None, is_remove_duplicate=False): + """ convert text-index into text-label. """ + result_list = [] + ignored_tokens = self.get_ignored_tokens() + [beg_idx, end_idx] = self.get_ignored_tokens() + batch_size = len(text_index) + for batch_idx in range(batch_size): + char_list = [] + conf_list = [] + for idx in range(len(text_index[batch_idx])): + if text_index[batch_idx][idx] in ignored_tokens: + continue + if int(text_index[batch_idx][idx]) == int(end_idx): + break + if is_remove_duplicate: + # only for predict + if idx > 0 and text_index[batch_idx][idx - 1] == text_index[ + batch_idx][idx]: + continue + char_list.append(self.character[int(text_index[batch_idx][ + idx])]) + if text_prob is not None: + conf_list.append(text_prob[batch_idx][idx]) + else: + conf_list.append(1) + text = ''.join(char_list) + result_list.append((text, np.mean(conf_list).tolist())) + return result_list + + def __call__(self, preds, label=None, *args, **kwargs): + """ + text = self.decode(text) + if label is None: + return text + else: + label = self.decode(label, is_remove_duplicate=False) + return text, label + """ + # if isinstance(preds, paddle.Tensor): + # preds = preds.numpy() + + preds_idx = preds.argmax(axis=2) + preds_prob = preds.max(axis=2) + text = self.decode(preds_idx, preds_prob, is_remove_duplicate=False) + if label is None: + return text + label = self.decode(label, is_remove_duplicate=False) + return text, label + + def get_ignored_tokens(self): + beg_idx = self.get_beg_end_flag_idx("beg") + end_idx = self.get_beg_end_flag_idx("end") + return [beg_idx, end_idx] + + def get_beg_end_flag_idx(self, beg_or_end): + if beg_or_end == "beg": + idx = np.array(self.dict[self.beg_str]) + elif beg_or_end == "end": + idx = np.array(self.dict[self.end_str]) + else: + assert False, "unsupport type %s in get_beg_end_flag_idx" \ + % beg_or_end + return idx + + +class RFLLabelDecode(BaseRecLabelDecode): + """ Convert between text-label and text-index """ + + def __init__(self, character_dict_path=None, use_space_char=False, + **kwargs): + super(RFLLabelDecode, self).__init__(character_dict_path, + use_space_char) + + def add_special_char(self, dict_character): + self.beg_str = "sos" + self.end_str = "eos" + dict_character = dict_character + dict_character = [self.beg_str] + dict_character + [self.end_str] + return dict_character + + def decode(self, text_index, text_prob=None, is_remove_duplicate=False): + """ convert text-index into text-label. """ + result_list = [] + ignored_tokens = self.get_ignored_tokens() + [beg_idx, end_idx] = self.get_ignored_tokens() + batch_size = len(text_index) + for batch_idx in range(batch_size): + char_list = [] + conf_list = [] + for idx in range(len(text_index[batch_idx])): + if text_index[batch_idx][idx] in ignored_tokens: + continue + if int(text_index[batch_idx][idx]) == int(end_idx): + break + if is_remove_duplicate: + # only for predict + if idx > 0 and text_index[batch_idx][idx - 1] == text_index[ + batch_idx][idx]: + continue + char_list.append(self.character[int(text_index[batch_idx][ + idx])]) + if text_prob is not None: + conf_list.append(text_prob[batch_idx][idx]) + else: + conf_list.append(1) + text = ''.join(char_list) + result_list.append((text, np.mean(conf_list).tolist())) + return result_list + + def __call__(self, preds, label=None, *args, **kwargs): + # if seq_outputs is not None: + if isinstance(preds, tuple) or isinstance(preds, list): + cnt_outputs, seq_outputs = preds + # if isinstance(seq_outputs, paddle.Tensor): + # seq_outputs = seq_outputs.numpy() + preds_idx = seq_outputs.argmax(axis=2) + preds_prob = seq_outputs.max(axis=2) + text = self.decode(preds_idx, preds_prob, is_remove_duplicate=False) + + if label is None: + return text + label = self.decode(label, is_remove_duplicate=False) + return text, label + + else: + cnt_outputs = preds + # if isinstance(cnt_outputs, paddle.Tensor): + # cnt_outputs = cnt_outputs.numpy() + cnt_length = [] + for lens in cnt_outputs: + length = round(np.sum(lens)) + cnt_length.append(length) + if label is None: + return cnt_length + label = self.decode(label, is_remove_duplicate=False) + length = [len(res[0]) for res in label] + return cnt_length, length + + def get_ignored_tokens(self): + beg_idx = self.get_beg_end_flag_idx("beg") + end_idx = self.get_beg_end_flag_idx("end") + return [beg_idx, end_idx] + + def get_beg_end_flag_idx(self, beg_or_end): + if beg_or_end == "beg": + idx = np.array(self.dict[self.beg_str]) + elif beg_or_end == "end": + idx = np.array(self.dict[self.end_str]) + else: + assert False, "unsupport type %s in get_beg_end_flag_idx" \ + % beg_or_end + return idx + + +class SEEDLabelDecode(BaseRecLabelDecode): + """ Convert between text-label and text-index """ + + def __init__(self, character_dict_path=None, use_space_char=False, + **kwargs): + super(SEEDLabelDecode, self).__init__(character_dict_path, + use_space_char) + + def add_special_char(self, dict_character): + self.padding_str = "padding" + self.end_str = "eos" + self.unknown = "unknown" + dict_character = dict_character + [ + self.end_str, self.padding_str, self.unknown + ] + return dict_character + + def get_ignored_tokens(self): + end_idx = self.get_beg_end_flag_idx("eos") + return [end_idx] + + def get_beg_end_flag_idx(self, beg_or_end): + if beg_or_end == "sos": + idx = np.array(self.dict[self.beg_str]) + elif beg_or_end == "eos": + idx = np.array(self.dict[self.end_str]) + else: + assert False, "unsupport type %s in get_beg_end_flag_idx" % beg_or_end + return idx + + def decode(self, text_index, text_prob=None, is_remove_duplicate=False): + """ convert text-index into text-label. """ + result_list = [] + [end_idx] = self.get_ignored_tokens() + batch_size = len(text_index) + for batch_idx in range(batch_size): + char_list = [] + conf_list = [] + for idx in range(len(text_index[batch_idx])): + if int(text_index[batch_idx][idx]) == int(end_idx): + break + if is_remove_duplicate: + # only for predict + if idx > 0 and text_index[batch_idx][idx - 1] == text_index[ + batch_idx][idx]: + continue + char_list.append(self.character[int(text_index[batch_idx][ + idx])]) + if text_prob is not None: + conf_list.append(text_prob[batch_idx][idx]) + else: + conf_list.append(1) + text = ''.join(char_list) + result_list.append((text, np.mean(conf_list).tolist())) + return result_list + + def __call__(self, preds, label=None, *args, **kwargs): + """ + text = self.decode(text) + if label is None: + return text + else: + label = self.decode(label, is_remove_duplicate=False) + return text, label + """ + preds_idx = preds["rec_pred"] + # if isinstance(preds_idx, paddle.Tensor): + # preds_idx = preds_idx.numpy() + if "rec_pred_scores" in preds: + preds_idx = preds["rec_pred"] + preds_prob = preds["rec_pred_scores"] + else: + preds_idx = preds["rec_pred"].argmax(axis=2) + preds_prob = preds["rec_pred"].max(axis=2) + text = self.decode(preds_idx, preds_prob, is_remove_duplicate=False) + if label is None: + return text + label = self.decode(label, is_remove_duplicate=False) + return text, label + + +class SRNLabelDecode(BaseRecLabelDecode): + """ Convert between text-label and text-index """ + + def __init__(self, character_dict_path=None, use_space_char=False, + **kwargs): + super(SRNLabelDecode, self).__init__(character_dict_path, + use_space_char) + self.max_text_length = kwargs.get('max_text_length', 25) + + def __call__(self, preds, label=None, *args, **kwargs): + pred = preds['predict'] + char_num = len(self.character_str) + 2 + # if isinstance(pred, paddle.Tensor): + # pred = pred.numpy() + pred = np.reshape(pred, [-1, char_num]) + + preds_idx = np.argmax(pred, axis=1) + preds_prob = np.max(pred, axis=1) + + preds_idx = np.reshape(preds_idx, [-1, self.max_text_length]) + + preds_prob = np.reshape(preds_prob, [-1, self.max_text_length]) + + text = self.decode(preds_idx, preds_prob) + + if label is None: + text = self.decode(preds_idx, preds_prob, is_remove_duplicate=False) + return text + label = self.decode(label) + return text, label + + def decode(self, text_index, text_prob=None, is_remove_duplicate=False): + """ convert text-index into text-label. """ + result_list = [] + ignored_tokens = self.get_ignored_tokens() + batch_size = len(text_index) + + for batch_idx in range(batch_size): + char_list = [] + conf_list = [] + for idx in range(len(text_index[batch_idx])): + if text_index[batch_idx][idx] in ignored_tokens: + continue + if is_remove_duplicate: + # only for predict + if idx > 0 and text_index[batch_idx][idx - 1] == text_index[ + batch_idx][idx]: + continue + char_list.append(self.character[int(text_index[batch_idx][ + idx])]) + if text_prob is not None: + conf_list.append(text_prob[batch_idx][idx]) + else: + conf_list.append(1) + + text = ''.join(char_list) + result_list.append((text, np.mean(conf_list).tolist())) + return result_list + + def add_special_char(self, dict_character): + dict_character = dict_character + [self.beg_str, self.end_str] + return dict_character + + def get_ignored_tokens(self): + beg_idx = self.get_beg_end_flag_idx("beg") + end_idx = self.get_beg_end_flag_idx("end") + return [beg_idx, end_idx] + + def get_beg_end_flag_idx(self, beg_or_end): + if beg_or_end == "beg": + idx = np.array(self.dict[self.beg_str]) + elif beg_or_end == "end": + idx = np.array(self.dict[self.end_str]) + else: + assert False, "unsupport type %s in get_beg_end_flag_idx" \ + % beg_or_end + return idx + + +class SARLabelDecode(BaseRecLabelDecode): + """ Convert between text-label and text-index """ + + def __init__(self, character_dict_path=None, use_space_char=False, + **kwargs): + super(SARLabelDecode, self).__init__(character_dict_path, + use_space_char) + + self.rm_symbol = kwargs.get('rm_symbol', False) + + def add_special_char(self, dict_character): + beg_end_str = "" + unknown_str = "" + padding_str = "" + dict_character = dict_character + [unknown_str] + self.unknown_idx = len(dict_character) - 1 + dict_character = dict_character + [beg_end_str] + self.start_idx = len(dict_character) - 1 + self.end_idx = len(dict_character) - 1 + dict_character = dict_character + [padding_str] + self.padding_idx = len(dict_character) - 1 + return dict_character + + def decode(self, text_index, text_prob=None, is_remove_duplicate=False): + """ convert text-index into text-label. """ + result_list = [] + ignored_tokens = self.get_ignored_tokens() + + batch_size = len(text_index) + for batch_idx in range(batch_size): + char_list = [] + conf_list = [] + for idx in range(len(text_index[batch_idx])): + if text_index[batch_idx][idx] in ignored_tokens: + continue + if int(text_index[batch_idx][idx]) == int(self.end_idx): + if text_prob is None and idx == 0: + continue + else: + break + if is_remove_duplicate: + # only for predict + if idx > 0 and text_index[batch_idx][idx - 1] == text_index[ + batch_idx][idx]: + continue + char_list.append(self.character[int(text_index[batch_idx][ + idx])]) + if text_prob is not None: + conf_list.append(text_prob[batch_idx][idx]) + else: + conf_list.append(1) + text = ''.join(char_list) + if self.rm_symbol: + comp = re.compile('[^A-Z^a-z^0-9^\u4e00-\u9fa5]') + text = text.lower() + text = comp.sub('', text) + result_list.append((text, np.mean(conf_list).tolist())) + return result_list + + def __call__(self, preds, label=None, *args, **kwargs): + # if isinstance(preds, paddle.Tensor): + # preds = preds.numpy() + preds_idx = preds.argmax(axis=2) + preds_prob = preds.max(axis=2) + + text = self.decode(preds_idx, preds_prob, is_remove_duplicate=False) + + if label is None: + return text + label = self.decode(label, is_remove_duplicate=False) + return text, label + + def get_ignored_tokens(self): + return [self.padding_idx] + + +class DistillationSARLabelDecode(SARLabelDecode): + """ + Convert + Convert between text-label and text-index + """ + + def __init__(self, + character_dict_path=None, + use_space_char=False, + model_name=["student"], + key=None, + multi_head=False, + **kwargs): + super(DistillationSARLabelDecode, self).__init__(character_dict_path, + use_space_char) + if not isinstance(model_name, list): + model_name = [model_name] + self.model_name = model_name + + self.key = key + self.multi_head = multi_head + + def __call__(self, preds, label=None, *args, **kwargs): + output = dict() + for name in self.model_name: + pred = preds[name] + if self.key is not None: + pred = pred[self.key] + if self.multi_head and isinstance(pred, dict): + pred = pred['sar'] + output[name] = super().__call__(pred, label=label, *args, **kwargs) + return output + + +class PRENLabelDecode(BaseRecLabelDecode): + """ Convert between text-label and text-index """ + + def __init__(self, character_dict_path=None, use_space_char=False, + **kwargs): + super(PRENLabelDecode, self).__init__(character_dict_path, + use_space_char) + + def add_special_char(self, dict_character): + padding_str = '' # 0 + end_str = '' # 1 + unknown_str = '' # 2 + + dict_character = [padding_str, end_str, unknown_str] + dict_character + self.padding_idx = 0 + self.end_idx = 1 + self.unknown_idx = 2 + + return dict_character + + def decode(self, text_index, text_prob=None): + """ convert text-index into text-label. """ + result_list = [] + batch_size = len(text_index) + + for batch_idx in range(batch_size): + char_list = [] + conf_list = [] + for idx in range(len(text_index[batch_idx])): + if text_index[batch_idx][idx] == self.end_idx: + break + if text_index[batch_idx][idx] in \ + [self.padding_idx, self.unknown_idx]: + continue + char_list.append(self.character[int(text_index[batch_idx][ + idx])]) + if text_prob is not None: + conf_list.append(text_prob[batch_idx][idx]) + else: + conf_list.append(1) + + text = ''.join(char_list) + if len(text) > 0: + result_list.append((text, np.mean(conf_list).tolist())) + else: + # here confidence of empty recog result is 1 + result_list.append(('', 1)) + return result_list + + def __call__(self, preds, label=None, *args, **kwargs): + # if isinstance(preds, paddle.Tensor): + # preds = preds.numpy() + preds_idx = preds.argmax(axis=2) + preds_prob = preds.max(axis=2) + text = self.decode(preds_idx, preds_prob) + if label is None: + return text + label = self.decode(label) + return text, label + + +class NRTRLabelDecode(BaseRecLabelDecode): + """ Convert between text-label and text-index """ + + def __init__(self, character_dict_path=None, use_space_char=True, **kwargs): + super(NRTRLabelDecode, self).__init__(character_dict_path, + use_space_char) + + def __call__(self, preds, label=None, *args, **kwargs): + + if len(preds) == 2: + preds_id = preds[0] + preds_prob = preds[1] + # if isinstance(preds_id, paddle.Tensor): + # preds_id = preds_id.numpy() + # if isinstance(preds_prob, paddle.Tensor): + # preds_prob = preds_prob.numpy() + if preds_id[0][0] == 2: + preds_idx = preds_id[:, 1:] + preds_prob = preds_prob[:, 1:] + else: + preds_idx = preds_id + text = self.decode(preds_idx, preds_prob, is_remove_duplicate=False) + if label is None: + return text + label = self.decode(label[:, 1:]) + else: + # if isinstance(preds, paddle.Tensor): + # preds = preds.numpy() + preds_idx = preds.argmax(axis=2) + preds_prob = preds.max(axis=2) + text = self.decode(preds_idx, preds_prob, is_remove_duplicate=False) + if label is None: + return text + label = self.decode(label[:, 1:]) + return text, label + + def add_special_char(self, dict_character): + dict_character = ['blank', '', '', ''] + dict_character + return dict_character + + def decode(self, text_index, text_prob=None, is_remove_duplicate=False): + """ convert text-index into text-label. """ + result_list = [] + batch_size = len(text_index) + for batch_idx in range(batch_size): + char_list = [] + conf_list = [] + for idx in range(len(text_index[batch_idx])): + try: + char_idx = self.character[int(text_index[batch_idx][idx])] + except: + continue + if char_idx == '': # end + break + char_list.append(char_idx) + if text_prob is not None: + conf_list.append(text_prob[batch_idx][idx]) + else: + conf_list.append(1) + text = ''.join(char_list) + result_list.append((text.lower(), np.mean(conf_list).tolist())) + return result_list + + +class ViTSTRLabelDecode(NRTRLabelDecode): + """ Convert between text-label and text-index """ + + def __init__(self, character_dict_path=None, use_space_char=False, + **kwargs): + super(ViTSTRLabelDecode, self).__init__(character_dict_path, + use_space_char) + + def __call__(self, preds, label=None, *args, **kwargs): + # if isinstance(preds, paddle.Tensor): + # preds = preds[:, 1:].numpy() + # else: + preds = preds[:, 1:] + preds_idx = preds.argmax(axis=2) + preds_prob = preds.max(axis=2) + text = self.decode(preds_idx, preds_prob, is_remove_duplicate=False) + if label is None: + return text + label = self.decode(label[:, 1:]) + return text, label + + def add_special_char(self, dict_character): + dict_character = ['', ''] + dict_character + return dict_character + + +class ABINetLabelDecode(NRTRLabelDecode): + """ Convert between text-label and text-index """ + + def __init__(self, character_dict_path=None, use_space_char=False, + **kwargs): + super(ABINetLabelDecode, self).__init__(character_dict_path, + use_space_char) + + def __call__(self, preds, label=None, *args, **kwargs): + if isinstance(preds, dict): + preds = preds['align'][-1].numpy() + # elif isinstance(preds, paddle.Tensor): + # preds = preds.numpy() + else: + preds = preds + + preds_idx = preds.argmax(axis=2) + preds_prob = preds.max(axis=2) + text = self.decode(preds_idx, preds_prob, is_remove_duplicate=False) + if label is None: + return text + label = self.decode(label) + return text, label + + def add_special_char(self, dict_character): + dict_character = [''] + dict_character + return dict_character + + +class SPINLabelDecode(AttnLabelDecode): + """ Convert between text-label and text-index """ + + def __init__(self, character_dict_path=None, use_space_char=False, + **kwargs): + super(SPINLabelDecode, self).__init__(character_dict_path, + use_space_char) + + def add_special_char(self, dict_character): + self.beg_str = "sos" + self.end_str = "eos" + dict_character = dict_character + dict_character = [self.beg_str] + [self.end_str] + dict_character + return dict_character + + +# class VLLabelDecode(BaseRecLabelDecode): +# """ Convert between text-label and text-index """ +# +# def __init__(self, character_dict_path=None, use_space_char=False, +# **kwargs): +# super(VLLabelDecode, self).__init__(character_dict_path, use_space_char) +# self.max_text_length = kwargs.get('max_text_length', 25) +# self.nclass = len(self.character) + 1 +# +# def decode(self, text_index, text_prob=None, is_remove_duplicate=False): +# """ convert text-index into text-label. """ +# result_list = [] +# ignored_tokens = self.get_ignored_tokens() +# batch_size = len(text_index) +# for batch_idx in range(batch_size): +# selection = np.ones(len(text_index[batch_idx]), dtype=bool) +# if is_remove_duplicate: +# selection[1:] = text_index[batch_idx][1:] != text_index[ +# batch_idx][:-1] +# for ignored_token in ignored_tokens: +# selection &= text_index[batch_idx] != ignored_token +# +# char_list = [ +# self.character[text_id - 1] +# for text_id in text_index[batch_idx][selection] +# ] +# if text_prob is not None: +# conf_list = text_prob[batch_idx][selection] +# else: +# conf_list = [1] * len(selection) +# if len(conf_list) == 0: +# conf_list = [0] +# +# text = ''.join(char_list) +# result_list.append((text, np.mean(conf_list).tolist())) +# return result_list +# +# def __call__(self, preds, label=None, length=None, *args, **kwargs): +# if len(preds) == 2: # eval mode +# text_pre, x = preds +# b = text_pre.shape[1] +# lenText = self.max_text_length +# nsteps = self.max_text_length +# +# if not isinstance(text_pre, paddle.Tensor): +# text_pre = paddle.to_tensor(text_pre, dtype='float32') +# +# out_res = paddle.zeros( +# shape=[lenText, b, self.nclass], dtype=x.dtype) +# out_length = paddle.zeros(shape=[b], dtype=x.dtype) +# now_step = 0 +# for _ in range(nsteps): +# if 0 in out_length and now_step < nsteps: +# tmp_result = text_pre[now_step, :, :] +# out_res[now_step] = tmp_result +# tmp_result = tmp_result.topk(1)[1].squeeze(axis=1) +# for j in range(b): +# if out_length[j] == 0 and tmp_result[j] == 0: +# out_length[j] = now_step + 1 +# now_step += 1 +# for j in range(0, b): +# if int(out_length[j]) == 0: +# out_length[j] = nsteps +# start = 0 +# output = paddle.zeros( +# shape=[int(out_length.sum()), self.nclass], dtype=x.dtype) +# for i in range(0, b): +# cur_length = int(out_length[i]) +# output[start:start + cur_length] = out_res[0:cur_length, i, :] +# start += cur_length +# net_out = output +# length = out_length +# +# else: # train mode +# net_out = preds[0] +# length = length +# net_out = paddle.concat([t[:l] for t, l in zip(net_out, length)]) +# text = [] +# if not isinstance(net_out, paddle.Tensor): +# net_out = paddle.to_tensor(net_out, dtype='float32') +# net_out = F.softmax(net_out, axis=1) +# for i in range(0, length.shape[0]): +# preds_idx = net_out[int(length[:i].sum()):int(length[:i].sum( +# ) + length[i])].topk(1)[1][:, 0].tolist() +# preds_text = ''.join([ +# self.character[idx - 1] +# if idx > 0 and idx <= len(self.character) else '' +# for idx in preds_idx +# ]) +# preds_prob = net_out[int(length[:i].sum()):int(length[:i].sum( +# ) + length[i])].topk(1)[0][:, 0] +# preds_prob = paddle.exp( +# paddle.log(preds_prob).sum() / (preds_prob.shape[0] + 1e-6)) +# text.append((preds_text, preds_prob.numpy()[0])) +# if label is None: +# return text +# label = self.decode(label) +# return text, label + + +class CANLabelDecode(BaseRecLabelDecode): + """ Convert between latex-symbol and symbol-index """ + + def __init__(self, character_dict_path=None, use_space_char=False, + **kwargs): + super(CANLabelDecode, self).__init__(character_dict_path, + use_space_char) + + def decode(self, text_index, preds_prob=None): + result_list = [] + batch_size = len(text_index) + for batch_idx in range(batch_size): + seq_end = text_index[batch_idx].argmin(0) + idx_list = text_index[batch_idx][:seq_end].tolist() + symbol_list = [self.character[idx] for idx in idx_list] + probs = [] + if preds_prob is not None: + probs = preds_prob[batch_idx][:len(symbol_list)].tolist() + + result_list.append([' '.join(symbol_list), probs]) + return result_list + + def __call__(self, preds, label=None, *args, **kwargs): + pred_prob, _, _, _ = preds + preds_idx = pred_prob.argmax(axis=2) + + text = self.decode(preds_idx) + if label is None: + return text + label = self.decode(label) + return text, label diff --git a/pp_onnx/utils.py b/pp_onnx/utils.py new file mode 100644 index 0000000..bb3a497 --- /dev/null +++ b/pp_onnx/utils.py @@ -0,0 +1,285 @@ +import numpy as np +import cv2 +import argparse +import math +from PIL import Image, ImageDraw, ImageFont + +# pathlib +from logzero import logger +from importlib.resources import files + +def get_rotate_crop_image(img, points): + ''' + img_height, img_width = img.shape[0:2] + left = int(np.min(points[:, 0])) + right = int(np.max(points[:, 0])) + top = int(np.min(points[:, 1])) + bottom = int(np.max(points[:, 1])) + img_crop = img[top:bottom, left:right, :].copy() + points[:, 0] = points[:, 0] - left + points[:, 1] = points[:, 1] - top + ''' + assert len(points) == 4, "shape of points must be 4*2" + img_crop_width = int( + max( + np.linalg.norm(points[0] - points[1]), + np.linalg.norm(points[2] - points[3]))) + img_crop_height = int( + max( + np.linalg.norm(points[0] - points[3]), + np.linalg.norm(points[1] - points[2]))) + pts_std = np.float32([[0, 0], [img_crop_width, 0], + [img_crop_width, img_crop_height], + [0, img_crop_height]]) + M = cv2.getPerspectiveTransform(points, pts_std) + dst_img = cv2.warpPerspective( + img, + M, (img_crop_width, img_crop_height), + borderMode=cv2.BORDER_REPLICATE, + flags=cv2.INTER_CUBIC) + dst_img_height, dst_img_width = dst_img.shape[0:2] + if dst_img_height * 1.0 / dst_img_width >= 1.5: + dst_img = np.rot90(dst_img) + return dst_img + +def get_minarea_rect_crop(img, points): + bounding_box = cv2.minAreaRect(np.array(points).astype(np.int32)) + points = sorted(list(cv2.boxPoints(bounding_box)), key=lambda x: x[0]) + + index_a, index_b, index_c, index_d = 0, 1, 2, 3 + if points[1][1] > points[0][1]: + index_a = 0 + index_d = 1 + else: + index_a = 1 + index_d = 0 + if points[3][1] > points[2][1]: + index_b = 2 + index_c = 3 + else: + index_b = 3 + index_c = 2 + + box = [points[index_a], points[index_b], points[index_c], points[index_d]] + crop_img = get_rotate_crop_image(img, np.array(box)) + return crop_img + + +def resize_img(img, input_size=600): + """ + resize img and limit the longest side of the image to input_size + """ + img = np.array(img) + im_shape = img.shape + im_size_max = np.max(im_shape[0:2]) + im_scale = float(input_size) / float(im_size_max) + img = cv2.resize(img, None, None, fx=im_scale, fy=im_scale) + return img + +def str_count(s): + """ + Count the number of Chinese characters, + a single English character and a single number + equal to half the length of Chinese characters. + args: + s(string): the input of string + return(int): + the number of Chinese characters + """ + import string + count_zh = count_pu = 0 + s_len = len(str(s)) + en_dg_count = 0 + for c in str(s): + if c in string.ascii_letters or c.isdigit() or c.isspace(): + en_dg_count += 1 + elif c.isalpha(): + count_zh += 1 + else: + count_pu += 1 + return s_len - math.ceil(en_dg_count / 2) + +def text_visual(texts, + scores, + img_h=400, + img_w=600, + threshold=0., + font_path="./fonts/simfang.ttf"): + """ + create new blank img and draw txt on it + args: + texts(list): the text will be draw + scores(list|None): corresponding score of each txt + img_h(int): the height of blank img + img_w(int): the width of blank img + font_path: the path of font which is used to draw text + return(array): + """ + if scores is not None: + assert len(texts) == len( + scores), "The number of txts and corresponding scores must match" + + def create_blank_img(): + blank_img = np.ones(shape=[img_h, img_w], dtype=np.uint8) * 255 + blank_img[:, img_w - 1:] = 0 + blank_img = Image.fromarray(blank_img).convert("RGB") + draw_txt = ImageDraw.Draw(blank_img) + return blank_img, draw_txt + + blank_img, draw_txt = create_blank_img() + + font_size = 20 + txt_color = (0, 0, 0) + # import IPython; IPython.embed(header='L-129') + font = ImageFont.truetype(str(font_path), font_size, encoding="utf-8") + + gap = font_size + 5 + txt_img_list = [] + count, index = 1, 0 + for idx, txt in enumerate(texts): + index += 1 + if scores[idx] < threshold or math.isnan(scores[idx]): + index -= 1 + continue + first_line = True + while str_count(txt) >= img_w // font_size - 4: + tmp = txt + txt = tmp[:img_w // font_size - 4] + if first_line: + new_txt = str(index) + ': ' + txt + first_line = False + else: + new_txt = ' ' + txt + draw_txt.text((0, gap * count), new_txt, txt_color, font=font) + txt = tmp[img_w // font_size - 4:] + if count >= img_h // gap - 1: + txt_img_list.append(np.array(blank_img)) + blank_img, draw_txt = create_blank_img() + count = 0 + count += 1 + if first_line: + new_txt = str(index) + ': ' + txt + ' ' + '%.3f' % (scores[idx]) + else: + new_txt = " " + txt + " " + '%.3f' % (scores[idx]) + draw_txt.text((0, gap * count), new_txt, txt_color, font=font) + # whether add new blank img or not + if count >= img_h // gap - 1 and idx + 1 < len(texts): + txt_img_list.append(np.array(blank_img)) + blank_img, draw_txt = create_blank_img() + count = 0 + count += 1 + txt_img_list.append(np.array(blank_img)) + if len(txt_img_list) == 1: + blank_img = np.array(txt_img_list[0]) + else: + blank_img = np.concatenate(txt_img_list, axis=1) + return np.array(blank_img) + +def draw_ocr(image, + boxes, + txts=None, + scores=None, + drop_score=0.5, + font_path=None): + """ + Visualize the results of OCR detection and recognition + args: + image(Image|array): RGB image + boxes(list): boxes with shape(N, 4, 2) + txts(list): the texts + scores(list): txxs corresponding scores + drop_score(float): only scores greater than drop_threshold will be visualized + font_path: the path of font which is used to draw text + return(array): + the visualized img + """ + if font_path is None: + SIMFANG_TTF = files('pp_onnx').joinpath('fonts/simfang.ttf') + font_path = SIMFANG_TTF + + if scores is None: + scores = [1] * len(boxes) + box_num = len(boxes) + for i in range(box_num): + if scores is not None and (scores[i] < drop_score or + math.isnan(scores[i])): + continue + box = np.reshape(np.array(boxes[i]), [-1, 1, 2]).astype(np.int64) + image = cv2.polylines(np.array(image), [box], True, (255, 0, 0), 2) + if txts is not None: + img = np.array(resize_img(image, input_size=600)) + txt_img = text_visual( + txts, + scores, + img_h=img.shape[0], + img_w=600, + threshold=drop_score, + font_path=font_path) + img = np.concatenate([np.array(img), np.array(txt_img)], axis=1) + return img + return image + +def base64_to_cv2(b64str): + import base64 + data = base64.b64decode(b64str.encode('utf8')) + data = np.frombuffer(data, np.uint8) + data = cv2.imdecode(data, cv2.IMREAD_COLOR) + return data + +def str2bool(v): + return v.lower() in ("true", "t", "1") + + + +def infer_args(): + parser = argparse.ArgumentParser() + + DET_MODEL_DIR = files('pp_onnx').joinpath('models/ch_PP-OCRv4/ch_PP-OCRv4_det_infer.onnx') + REC_MODEL_DIR = files('pp_onnx').joinpath('models/ch_PP-OCRv4/ch_PP-OCRv4_rec_infer.onnx') + PPOCR_KEYS_V1 = files('pp_onnx').joinpath('models/ch_ppocr_server_v2.0/ppocr_keys_v1.txt') + SIMFANG_TTF = files('pp_onnx').joinpath('fonts/simfang.ttf') + CLS_MODEL_DIR = files('pp_onnx').joinpath('models/ch_ppocr_server_v2.0/cls/cls.onnx') + + # params for text detector + parser.add_argument("--image_dir", type=str) + parser.add_argument("--page_num", type=int, default=0) + parser.add_argument("--det_algorithm", type=str, default='DB') + parser.add_argument("--det_model_dir", type=str, default=DET_MODEL_DIR) + parser.add_argument("--det_limit_side_len", type=float, default=960) + parser.add_argument("--det_limit_type", type=str, default='max') + parser.add_argument("--det_box_type", type=str, default='quad') + + # DB parmas + parser.add_argument("--det_db_thresh", type=float, default=0.3) + parser.add_argument("--det_db_box_thresh", type=float, default=0.6) + parser.add_argument("--det_db_unclip_ratio", type=float, default=1.5) + parser.add_argument("--max_batch_size", type=int, default=10) + parser.add_argument("--use_dilation", type=str2bool, default=False) + parser.add_argument("--det_db_score_mode", type=str, default="fast") + + # params for text recognizer + parser.add_argument("--rec_algorithm", type=str, default='SVTR_LCNet') + parser.add_argument("--rec_model_dir", type=str, default=REC_MODEL_DIR) + parser.add_argument("--rec_image_inverse", type=str2bool, default=True) + parser.add_argument("--rec_image_shape", type=str, default="3, 32, 320") + parser.add_argument("--rec_batch_num", type=int, default=6) + parser.add_argument("--max_text_length", type=int, default=25) + parser.add_argument( "--rec_char_dict_path", type=str, default=PPOCR_KEYS_V1) + parser.add_argument("--use_space_char", type=str2bool, default=True) + parser.add_argument( "--vis_font_path", type=str, default=SIMFANG_TTF) + parser.add_argument("--drop_score", type=float, default=0.5) + + # params for text classifier + parser.add_argument("--use_angle_cls", type=str2bool, default=False) + parser.add_argument("--cls_model_dir", type=str, default=CLS_MODEL_DIR) + parser.add_argument("--cls_image_shape", type=str, default="3, 48, 192") + parser.add_argument("--label_list", type=list, default=['0', '180']) + parser.add_argument("--cls_batch_num", type=int, default=6) + parser.add_argument("--cls_thresh", type=float, default=0.9) + + # others + parser.add_argument("--save_crop_res", type=str2bool, default=False) + # parser.add_argument( "--draw_img_save_dir", type=str, default="./onnx/inference_results") + # parser.add_argument("--crop_res_save_dir", type=str, default="./onnx/output") + + return parser \ No newline at end of file diff --git a/requirements.txt b/requirements.txt new file mode 100644 index 0000000..b0415aa --- /dev/null +++ b/requirements.txt @@ -0,0 +1,32 @@ +colorama==0.4.6 +coloredlogs==15.0.1 +cycler==0.11.0 +flatbuffers==23.5.26 +fonttools==4.38.0 +humanfriendly==10.0 +imageio==2.31.1 +imgaug==0.4.0 +kiwisolver==1.4.4 +lmdb==1.4.1 +matplotlib==3.5.3 +mpmath==1.3.0 +networkx==2.6.3 +numpy==1.21.6 +onnxruntime==1.14.1 +opencv-python==3.4.18.65 +packaging==23.1 +Pillow==9.5.0 +protobuf==4.23.4 +pyclipper==1.3.0.post4 +pyparsing==3.1.0 +pyreadline==2.1 +python-dateutil==2.8.2 +PyWavelets==1.3.0 +scikit-image==0.19.3 +scipy==1.7.3 +shapely==2.0.1 +six==1.16.0 +sympy==1.10.1 +tifffile==2021.11.2 +tqdm==4.65.0 +typing_extensions==4.7.1 diff --git a/result_img/1.jpg b/result_img/1.jpg new file mode 100644 index 0000000..a9a906e Binary files /dev/null and b/result_img/1.jpg differ diff --git a/result_img/3.jpg b/result_img/3.jpg new file mode 100644 index 0000000..d263334 Binary files /dev/null and b/result_img/3.jpg differ diff --git a/result_img/draw_ocr.jpg b/result_img/draw_ocr.jpg new file mode 100644 index 0000000..12d4ea1 Binary files /dev/null and b/result_img/draw_ocr.jpg differ diff --git a/result_img/draw_ocr2.jpg b/result_img/draw_ocr2.jpg new file mode 100644 index 0000000..9ed009e Binary files /dev/null and b/result_img/draw_ocr2.jpg differ diff --git a/result_img/draw_ocr3.jpg b/result_img/draw_ocr3.jpg new file mode 100644 index 0000000..72742cd Binary files /dev/null and b/result_img/draw_ocr3.jpg differ diff --git a/result_img/draw_ocr4.jpg b/result_img/draw_ocr4.jpg new file mode 100644 index 0000000..4f07a9c Binary files /dev/null and b/result_img/draw_ocr4.jpg differ diff --git a/result_img/draw_ocr5.jpg b/result_img/draw_ocr5.jpg new file mode 100644 index 0000000..5f3cee9 Binary files /dev/null and b/result_img/draw_ocr5.jpg differ diff --git a/result_img/draw_ocr_1.jpg b/result_img/draw_ocr_1.jpg new file mode 100644 index 0000000..ac37e37 Binary files /dev/null and b/result_img/draw_ocr_1.jpg differ diff --git a/result_img/draw_ocr_996.jpg b/result_img/draw_ocr_996.jpg new file mode 100644 index 0000000..ec18faa Binary files /dev/null and b/result_img/draw_ocr_996.jpg differ diff --git a/result_img/draw_ocr_996_1.jpg b/result_img/draw_ocr_996_1.jpg new file mode 100644 index 0000000..d263334 Binary files /dev/null and b/result_img/draw_ocr_996_1.jpg differ diff --git a/result_img/rec.onnx b/result_img/rec.onnx new file mode 100644 index 0000000..d61384e Binary files /dev/null and b/result_img/rec.onnx differ diff --git a/test_img/1.jpg b/test_img/1.jpg new file mode 100644 index 0000000..0801017 Binary files /dev/null and b/test_img/1.jpg differ diff --git a/test_img/10.jpg b/test_img/10.jpg new file mode 100644 index 0000000..698e0df Binary files /dev/null and b/test_img/10.jpg differ diff --git a/test_img/11.jpg b/test_img/11.jpg new file mode 100644 index 0000000..82a45c4 Binary files /dev/null and b/test_img/11.jpg differ diff --git a/test_img/12.jpg b/test_img/12.jpg new file mode 100644 index 0000000..0b5b656 Binary files /dev/null and b/test_img/12.jpg differ diff --git a/test_img/13.jpg b/test_img/13.jpg new file mode 100644 index 0000000..1f832d7 Binary files /dev/null and b/test_img/13.jpg differ diff --git a/test_img/14.jpg b/test_img/14.jpg new file mode 100644 index 0000000..2aae5f7 Binary files /dev/null and b/test_img/14.jpg differ diff --git a/test_img/15.jpg b/test_img/15.jpg new file mode 100644 index 0000000..e278adf Binary files /dev/null and b/test_img/15.jpg differ diff --git a/test_img/16.jpg b/test_img/16.jpg new file mode 100644 index 0000000..0c3cc4d Binary files /dev/null and b/test_img/16.jpg differ diff --git a/test_img/17.jpg b/test_img/17.jpg new file mode 100644 index 0000000..050bf7c Binary files /dev/null and b/test_img/17.jpg differ diff --git a/test_img/18.jpg b/test_img/18.jpg new file mode 100644 index 0000000..b922702 Binary files /dev/null and b/test_img/18.jpg differ diff --git a/test_img/19.jpg b/test_img/19.jpg new file mode 100644 index 0000000..ffcb2e4 Binary files /dev/null and b/test_img/19.jpg differ diff --git a/test_img/2.jpg b/test_img/2.jpg new file mode 100644 index 0000000..2656ad0 Binary files /dev/null and b/test_img/2.jpg differ diff --git a/test_img/20.jpg b/test_img/20.jpg new file mode 100644 index 0000000..0353874 Binary files /dev/null and b/test_img/20.jpg differ diff --git a/test_img/21.jpg b/test_img/21.jpg new file mode 100644 index 0000000..4259c04 Binary files /dev/null and b/test_img/21.jpg differ diff --git a/test_img/22.png b/test_img/22.png new file mode 100644 index 0000000..4906b27 Binary files /dev/null and b/test_img/22.png differ diff --git a/test_img/3.jpg b/test_img/3.jpg new file mode 100644 index 0000000..30ea9cd Binary files /dev/null and b/test_img/3.jpg differ diff --git a/test_img/4.jpg b/test_img/4.jpg new file mode 100644 index 0000000..ed91b8c Binary files /dev/null and b/test_img/4.jpg differ diff --git a/test_img/5.jpg b/test_img/5.jpg new file mode 100644 index 0000000..3cd2e79 Binary files /dev/null and b/test_img/5.jpg differ diff --git a/test_img/6.jpg b/test_img/6.jpg new file mode 100644 index 0000000..5c3329a Binary files /dev/null and b/test_img/6.jpg differ diff --git a/test_img/7.jpg b/test_img/7.jpg new file mode 100644 index 0000000..448d0f1 Binary files /dev/null and b/test_img/7.jpg differ diff --git a/test_img/8.jpg b/test_img/8.jpg new file mode 100644 index 0000000..9d6aaee Binary files /dev/null and b/test_img/8.jpg differ diff --git a/test_img/9.jpg b/test_img/9.jpg new file mode 100644 index 0000000..e768d8a Binary files /dev/null and b/test_img/9.jpg differ diff --git a/test_ocr.py b/test_ocr.py new file mode 100644 index 0000000..09ac674 --- /dev/null +++ b/test_ocr.py @@ -0,0 +1,45 @@ +import cv2 +import time +from pp_onnx.onnx_paddleocr import ONNXPaddleOcr, draw_ocr + +# 优化参数配置，提高速度 +model = ONNXPaddleOcr( + use_angle_cls=True, + use_gpu=False, + providers=['AzureExecutionProvider'], + provider_options=[{'device_id': 0}], + det_limit_type='max', +) + +def sav2Img(org_img, result, name="./result_img/3.jpg"): + from PIL import Image + result = result[0] + image = org_img[:, :, ::-1] + boxes = [line[0] for line in result] + txts = [line[1][0] for line in result] + scores = [line[1][1] for line in result] + im_show = draw_ocr(image, boxes, txts, scores) + im_show = Image.fromarray(im_show) + im_show.save(name) + + +# 执行OCR推理 +img = cv2.imread('./test_img/3.jpg') +if img is None: + print(f"❌ 未找到图像文件") +else: + # 图像预处理 - 缩小图像尺寸加速处理 + h, w = img.shape[:2] + max_size = 1080 + if max(h, w) > max_size: + scale = max_size / max(h, w) + img = cv2.resize(img, (int(w * scale), int(h * scale))) + + s = time.time() + result = model.ocr(img) + e = time.time() + print(f"total time: {e - s:.3f} 秒") + print("result:", result) + for box in result[0]: + print(box) + sav2Img(img, result) \ No newline at end of file