face-recognition示例程序检测到人脸时崩溃问题求助

运行face-recognition库的官方示例程序时，直播视频可正常显示，但人脸进入画面后程序立即崩溃，Windows 10和Ubuntu系统测试均出现相同错误。

示例代码 #

import face_recognition
import cv2
import numpy as np

# 获取默认摄像头（编号0）的引用
video_capture = cv2.VideoCapture(0)

# 加载样本图片并学习识别
obama_image = face_recognition.load_image_file("obama.jpg")
obama_face_encoding = face_recognition.face_encodings(obama_image)[0]

# 加载第二张样本图片并学习识别
biden_image = face_recognition.load_image_file("biden.jpg")
biden_face_encoding = face_recognition.face_encodings(biden_image)[0]

# 创建已知人脸编码和对应名称的数组
known_face_encodings = [
    obama_face_encoding,
    biden_face_encoding
]
known_face_names = [
    "Barack Obama",
    "Joe Biden"
]

# 初始化变量
face_locations = []
face_encodings = []
face_names = []
process_this_frame = True

while True:
    # 获取一帧视频
    ret, frame = video_capture.read()

    # 每隔一帧处理一次以节省时间
    if process_this_frame:
        # 将视频帧缩小到1/4尺寸，加快人脸识别速度
        small_frame = cv2.resize(frame, (0, 0), fx=0.25, fy=0.25)
        # 将OpenCV的BGR格式转换为face-recognition需要的RGB格式
        rgb_small_frame = small_frame[:, :, ::-1]
        # 查找当前视频帧中的所有人脸和人脸编码
        face_locations = face_recognition.face_locations(rgb_small_frame)
        face_encodings = face_recognition.face_encodings(rgb_small_frame, face_locations)
        face_names = []
        for face_encoding in face_encodings:
            # 检查人脸是否与已知人脸匹配
            matches = face_recognition.compare_faces(known_face_encodings, face_encoding)
            name = "Unknown"

            # 选择与当前人脸距离最小的已知人脸作为最佳匹配
            face_distances = face_recognition.face_distance(known_face_encodings, face_encoding)
            best_match_index = np.argmin(face_distances)
            if matches[best_match_index]:
                name = known_face_names[best_match_index]

            face_names.append(name)

    process_this_frame = not process_this_frame

    # 显示识别结果
    for (top, right, bottom, left), name in zip(face_locations, face_names):
        # 将人脸位置缩放回原帧尺寸
        top *= 4
        right *= 4
        bottom *= 4
        left *= 4

        # 绘制人脸边框
        cv2.rectangle(frame, (left, top), (right, bottom), (0, 0, 255), 2)

        # 绘制姓名标签
        cv2.rectangle(frame, (left, bottom - 35), (right, bottom), (0, 0, 255), cv2.FILLED)
        font = cv2.FONT_HERSHEY_DUPLEX
        cv2.putText(frame, name, (left + 6, bottom - 6), font, 1.0, (255, 255, 255), 1)

    # 显示视频画面
    cv2.imshow('Video', frame)

    # 按下q键退出程序
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

# 释放摄像头资源
video_capture.release()
cv2.destroyAllWindows()

报错信息 #

Traceback (most recent call last):
  File ".\main.py", line 55, in <module>
    face_encodings = face_recognition.face_encodings(rgb_small_frame, face_loca
  File "C:\Users\markc\AppData\Local\Programs\Python\Python38\lib\site-packages
    return [np.array(face_encoder.compute_face_descriptor(face_image, raw_landm
  File "C:\Users\markc\AppData\Local\Programs\Python\Python38\lib\site-packages
    return [np.array(face_encoder.compute_face_descriptor(face_image, raw_landm
TypeError: compute_face_descriptor(): incompatible function arguments. The foll
    1. (self: _dlib_pybind11.face_recognition_model_v1, img: numpy.ndarray[(rowjitters: int = 0, padding: float = 0.25) -> _dlib_pybind11.vector
    2. (self: _dlib_pybind11.face_recognition_model_v1, img: numpy.ndarray[(row
    3. (self: _dlib_pybind11.face_recognition_model_v1, img: numpy.ndarray[(rowm_jitters: int = 0, padding: float = 0.25) -> _dlib_pybind11.vectors
    4. (self: _dlib_pybind11.face_recognition_model_v1, batch_img: List[numpy.nl_object_detections], num_jitters: int = 0, padding: float = 0.25) -> _dlib_pyb
    5. (self: _dlib_pybind11.face_recognition_model_v1, batch_img: List[numpy.nd11.vectors

已安装库版本 #

• dlib==19.24.1
• face-recognition==1.3.0
• face-recognition-models==0.3.0
• numpy==1.24.2
• opencv-python==4.7.0.72

解决方案 #

核心问题是版本不兼容：numpy 1.24+移除了旧版dlib/face-recognition依赖的API，导致compute_face_descriptor调用参数不匹配。

方法1：降级numpy到兼容版本 #

执行以下命令降级numpy，这是最稳妥的解决方式：

pip install numpy==1.23.5

方法2：升级face-recognition到最新版本（可选） #

如果不想降级numpy，可以尝试升级face-recognition到最新版（需确保dlib版本同步兼容，部分环境可能需要重新编译dlib）：

pip install --upgrade face-recognition

内容的提问来源于stack exchange，提问作者MichaelLong102