-
RetinaFace推理代码在PyTorch中实现,采用MobileNet作为后端的版本:retinaface-pytorch
资源介绍
PyTorch中带有MobileNet后端的RetinaFace推理代码
步骤1:
cd cython
python setup.py build_ext --inplace
第2步:
python inference.py
评估(宽屏):
Easy Val AP:0.8872715908531869 中值AP:0.8663337842229522 硬值AP:0.771796729363941
试验结果:
参考:
@inproceedings {deng2019retinaface,标题= {RetinaFace:野外单阶段密集脸定位},作者= {Deng,Jiankang和Guo,Jia和Yuxiang,Zhou和Jinke Yu和Irene Kotsia和Zafeiriou,Stefanos},书名= { arxiv},年份= {2019}}