目前可以成功动态更换模型运行的
This commit is contained in:
@ -4,6 +4,11 @@ import insightface
|
||||
from insightface.app import FaceAnalysis
|
||||
from io import BytesIO
|
||||
from PIL import Image
|
||||
import logging
|
||||
|
||||
# 配置日志（便于排查）
|
||||
logging.basicConfig(level=logging.INFO, format='%(asctime)s - [FaceUtil] - %(levelname)s - %(message)s')
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
# 全局变量存储InsightFace引擎和特征列表
|
||||
_insightface_app = None
|
||||
@ -11,135 +16,141 @@ _feature_list = []
|
||||
|
||||
|
||||
def init_insightface():
|
||||
"""初始化InsightFace引擎"""
|
||||
"""初始化InsightFace引擎（确保成功后再使用）"""
|
||||
global _insightface_app
|
||||
try:
|
||||
print("正在初始化InsightFace引擎...")
|
||||
app = FaceAnalysis(name='buffalo_l', root='~/.insightface')
|
||||
app.prepare(ctx_id=0, det_size=(640, 640))
|
||||
print("InsightFace引擎初始化完成")
|
||||
if _insightface_app is not None:
|
||||
logger.info("InsightFace引擎已初始化，无需重复执行")
|
||||
return _insightface_app
|
||||
|
||||
logger.info("正在初始化InsightFace引擎（模型：buffalo_l）...")
|
||||
# 手动指定模型下载路径（避免权限问题，可选）
|
||||
app = FaceAnalysis(
|
||||
name='buffalo_l',
|
||||
root='~/.insightface', # 模型默认下载路径
|
||||
providers=['CPUExecutionProvider'] # 强制用CPU（若有GPU可加'CUDAExecutionProvider'）
|
||||
)
|
||||
app.prepare(ctx_id=0, det_size=(640, 640)) # det_size越大，小人脸检测越准
|
||||
logger.info("InsightFace引擎初始化完成")
|
||||
_insightface_app = app
|
||||
return app
|
||||
except Exception as e:
|
||||
print(f"InsightFace初始化失败: {e}")
|
||||
logger.error(f"InsightFace初始化失败：{str(e)}", exc_info=True) # 打印详细堆栈
|
||||
_insightface_app = None
|
||||
return None
|
||||
|
||||
|
||||
def add_binary_data(binary_data):
|
||||
"""
|
||||
接收单张图片的二进制数据、提取特征并保存
|
||||
|
||||
参数:
|
||||
binary_data: 图片的二进制数据（bytes类型）
|
||||
|
||||
返回:
|
||||
成功提取特征时返回 (True, 特征值numpy数组)
|
||||
失败时返回 (False, None)
|
||||
返回：(True, 特征值numpy数组) 或 (False, 错误信息字符串)
|
||||
"""
|
||||
global _insightface_app, _feature_list
|
||||
|
||||
# 1. 先检查引擎是否初始化成功
|
||||
if not _insightface_app:
|
||||
print("引擎未初始化、无法处理")
|
||||
return False, None
|
||||
init_result = init_insightface() # 尝试重新初始化
|
||||
if not init_result:
|
||||
error_msg = "InsightFace引擎未初始化，无法检测人脸"
|
||||
logger.error(error_msg)
|
||||
return False, error_msg
|
||||
|
||||
try:
|
||||
# 直接处理二进制数据: 转换为图像格式
|
||||
img = Image.open(BytesIO(binary_data))
|
||||
frame = cv2.cvtColor(np.array(img), cv2.COLOR_RGB2BGR)
|
||||
# 2. 验证二进制数据有效性
|
||||
if len(binary_data) < 1024: # 过滤过小的无效图片（小于1KB）
|
||||
error_msg = f"图片过小（{len(binary_data)}字节），可能不是有效图片"
|
||||
logger.warning(error_msg)
|
||||
return False, error_msg
|
||||
|
||||
# 提取特征
|
||||
# 3. 二进制数据转CV2格式（关键步骤，避免通道错误）
|
||||
try:
|
||||
img = Image.open(BytesIO(binary_data)).convert("RGB") # 强制转RGB
|
||||
frame = cv2.cvtColor(np.array(img), cv2.COLOR_RGB2BGR) # InsightFace需要BGR格式
|
||||
except Exception as e:
|
||||
error_msg = f"图片格式转换失败：{str(e)}"
|
||||
logger.error(error_msg, exc_info=True)
|
||||
return False, error_msg
|
||||
|
||||
# 4. 检查图片尺寸（避免极端尺寸导致检测失败）
|
||||
height, width = frame.shape[:2]
|
||||
if height < 64 or width < 64: # 人脸检测最小建议尺寸
|
||||
error_msg = f"图片尺寸过小（{width}x{height}），需至少64x64像素"
|
||||
logger.warning(error_msg)
|
||||
return False, error_msg
|
||||
|
||||
# 5. 调用InsightFace检测人脸
|
||||
logger.info(f"开始检测人脸（图片尺寸：{width}x{height}，格式：BGR）")
|
||||
faces = _insightface_app.get(frame)
|
||||
if faces:
|
||||
# 获取当前提取的特征值
|
||||
current_feature = faces[0].embedding
|
||||
# 添加到特征列表
|
||||
_feature_list.append(current_feature)
|
||||
print(f"已累计 {len(_feature_list)} 个特征")
|
||||
# 返回成功标志和当前特征值
|
||||
return True, current_feature
|
||||
else:
|
||||
print("二进制数据中未检测到人脸")
|
||||
return False, None
|
||||
|
||||
if not faces:
|
||||
error_msg = "未检测到人脸（请确保图片包含清晰正面人脸，无遮挡、不模糊）"
|
||||
logger.warning(error_msg)
|
||||
return False, error_msg
|
||||
|
||||
# 6. 提取特征并保存
|
||||
current_feature = faces[0].embedding
|
||||
_feature_list.append(current_feature)
|
||||
logger.info(f"人脸检测成功，提取特征值（维度：{current_feature.shape[0]}），累计特征数：{len(_feature_list)}")
|
||||
return True, current_feature
|
||||
|
||||
except Exception as e:
|
||||
print(f"处理二进制数据出错: {e}")
|
||||
return False, None
|
||||
error_msg = f"处理图片时发生异常：{str(e)}"
|
||||
logger.error(error_msg, exc_info=True)
|
||||
return False, error_msg
|
||||
|
||||
|
||||
# 以下函数保持不变（get_average_feature/clear_features/get_feature_list）
|
||||
def get_average_feature(features=None):
|
||||
"""
|
||||
计算多个特征向量的平均值
|
||||
|
||||
参数:
|
||||
features: 可选、特征值列表。如果未提供、则使用全局存储的_feature_list
|
||||
每个元素可以是字符串格式或numpy数组
|
||||
|
||||
返回:
|
||||
单一平均特征向量的numpy数组、若无可计算数据则返回None
|
||||
"""
|
||||
global _feature_list
|
||||
|
||||
# 如果未提供features参数、则使用全局特征列表
|
||||
if features is None:
|
||||
features = _feature_list
|
||||
|
||||
try:
|
||||
# 验证输入是否为列表且不为空
|
||||
if features is None:
|
||||
features = _feature_list
|
||||
if not isinstance(features, list) or len(features) == 0:
|
||||
print("输入必须是包含至少一个特征值的列表")
|
||||
logger.warning("输入必须是包含至少一个特征值的列表")
|
||||
return None
|
||||
|
||||
# 处理每个特征值
|
||||
processed_features = []
|
||||
for i, embedding in enumerate(features):
|
||||
try:
|
||||
if isinstance(embedding, str):
|
||||
# 处理包含括号和逗号的字符串格式
|
||||
embedding_str = embedding.replace('[', '').replace(']', '').replace(',', ' ').strip()
|
||||
embedding_list = [float(num) for num in embedding_str.split() if num.strip()]
|
||||
embedding_np = np.array(embedding_list, dtype=np.float32)
|
||||
else:
|
||||
embedding_np = np.array(embedding, dtype=np.float32)
|
||||
|
||||
# 验证特征值格式
|
||||
if len(embedding_np.shape) == 1:
|
||||
processed_features.append(embedding_np)
|
||||
print(f"已添加第 {i + 1} 个特征值用于计算平均值")
|
||||
logger.info(f"已添加第 {i + 1} 个特征值用于计算平均值")
|
||||
else:
|
||||
print(f"跳过第 {i + 1} 个特征值、不是一维数组")
|
||||
|
||||
logger.warning(f"跳过第 {i + 1} 个特征值：不是一维数组")
|
||||
except Exception as e:
|
||||
print(f"处理第 {i + 1} 个特征值时出错: {e}")
|
||||
logger.error(f"处理第 {i + 1} 个特征值时出错：{str(e)}")
|
||||
|
||||
# 确保有有效的特征值
|
||||
if not processed_features:
|
||||
print("没有有效的特征值用于计算平均值")
|
||||
logger.warning("没有有效的特征值用于计算平均值")
|
||||
return None
|
||||
|
||||
# 检查所有特征向量维度是否相同
|
||||
dims = {feat.shape[0] for feat in processed_features}
|
||||
if len(dims) > 1:
|
||||
print(f"特征值维度不一致、无法计算平均值。检测到的维度: {dims}")
|
||||
logger.error(f"特征值维度不一致：{dims}，无法计算平均值")
|
||||
return None
|
||||
|
||||
# 计算平均值
|
||||
avg_feature = np.mean(processed_features, axis=0)
|
||||
print(f"成功计算 {len(processed_features)} 个特征值的平均特征向量、维度: {avg_feature.shape[0]}")
|
||||
|
||||
logger.info(f"计算成功：{len(processed_features)} 个特征值的平均向量（维度：{avg_feature.shape[0]}）")
|
||||
return avg_feature
|
||||
|
||||
except Exception as e:
|
||||
print(f"计算平均特征值时出错: {e}")
|
||||
logger.error(f"计算平均特征值出错：{str(e)}", exc_info=True)
|
||||
return None
|
||||
|
||||
|
||||
def clear_features():
|
||||
"""清空已存储的特征数据"""
|
||||
global _feature_list
|
||||
_feature_list = []
|
||||
print("已清空所有特征数据")
|
||||
logger.info("已清空所有特征数据")
|
||||
|
||||
|
||||
def get_feature_list():
|
||||
"""获取当前存储的特征列表"""
|
||||
global _feature_list
|
||||
return _feature_list.copy() # 返回副本防止外部直接修改
|
||||
logger.info(f"当前特征列表长度：{len(_feature_list)}")
|
||||
return _feature_list.copy()
|
||||
Reference in New Issue
Block a user