有关retinaface模型

代码链接：GitHub - serengil/retinaface: RetinaFace: Deep Face Detection Library for Python

请问直接使用faces = RetinaFace.detect_faces(frame)来处理的速度为什么这么慢呀？加载的速度慢的离谱了都。

已经得到解决，请使用封装好的库进行检测。

retinaface · PyPI

 如上↑,detector.predict(rgb_image)返回的是一个字典，自行按键值取用即可。

#pip3 install opencv-python
import cv2 
from retinaface import RetinaFace

# init with 'normal' accuracy option (resize width or height to 800 )
# or you can choice 'speed' (resize to 320)
# or you can initiate with no parameter for running with original image size
detector = RetinaFace(quality="normal")

# same with cv2.imread,cv2.cvtColor 
rgb_image = detector.read("data/hian.jpg")

faces = detector.predict(rgb_image)
# faces is list of face dictionary
# each face dictionary contains x1 y1 x2 y2 left_eye right_eye nose left_lip right_lip
# faces=[{"x1":20,"y1":32, ... }, ...]

result_img = detector.draw(rgb_image,faces)

# save ([...,::-1] : rgb -> bgr )
cv2.imwrite("data/result_img.jpg",result_img[...,::-1])

# show using cv2
# cv2.imshow("result",result_img[...,::-1])
# cv2.waitKey()