ShiqiYu/libfacedetection.train

训练完成导出C++部署不成功

nianfd opened this issue · 0 comments

nianfd commented

训练正常
python tools/test_widerface.py 也正常

  1. python tools/detect_image.py ./configs/yunet_n.py ./weights/yunet_n.pth ./image.jpg
    检测的人脸位置正常,无关键点画出

  2. python tools/yunet2cpp.py ./configs/yunet_n.py ./weights/yunet_n.pth 转换,成功,但C++运行时出现以下错误:
    bbox_decode: The bbox dim must be 4.
    bbox_decode: The bbox dim must be 4.
    bbox_decode: The bbox dim must be 4.
    kps_decode: The kps dim must be even.

直接使用./weights里您提供的预训练[yunet_n.pth]出现上述同样的错误