yxqAIxp comments

Results 38 comments of


                                            yxqAIxp

You can try [PFLD_68points_Pytorch](https://github.com/github-luffy/PFLD_68points_Pytorch)

hi, code:c = w * (1.0 - math.log(1.0 + w/epsilon)),math.log is equal to ln, that is no problem.

PFLD uses the regression loss function, and the cross entropy loss of the classification is not used, and there is no confidence score.

To output the key point heat map, you can look at the papers and projects of cpm (convolutional pose machine)

I have got the same problem.But after onnx->trt， trt result is right！ I am so confused.

> 作者你好，我要在30显卡上跑，cuda得11以上，其他的python包需要什么版本的呢？ python3版本就可以

这样方便加入其他数据集数据增加多样性可以保证提升效果

> 为了加掩码具体什么意思，加掩码是什么操作，作用是什么