Shiwei Hou

@alexanderswerdlow I only tried this on cpu, don't know how it compatiable with TPU.

Is it score you got it ? I find the visual-performance is better than score-performance. Maybe the `pose_decode` file need to be update. But i'm not sure what's wrong with...

I used this evaluation function just copying from [ai-challenger](https://github.com/AIChallenger/AI_Challenger_2017/blob/master/Evaluation/keypoint_eval/keypoint_eval.py). Maybe the file `model_json.py` has some wrong. I'll check it and hope that you can also find resolution sooner.

@LeifSun Thanks for your answer. Maybe guys @Henrietta95 and other people can tried his solution, i have no enough time to do this, sorry for that.

@LeifSun hi, could you tell me where have you changed about `depthwise separable convolution`?

hi, @gavians ,maybe they removed the dataset, you can try COCO dataset and remember to change model input & output to suitable for COCO format.

我的很小啊, 都是在0.01~0.001范围内, 源码用了mse和smooth l1 loss, 数值不可能大的 amitabhama 于2019年11月29日周五 上午11:46写道： > 我用的pytorch, 用您源码训练loss也很大, 2w以上 > > — > You are receiving this because you are subscribed to this thread. > Reply...

@amitabhama loss具体数值不重要, 重要的是 1. 你的loss计算方式是什么样子, loss大小应该有个大概的, 要确保的是这个loss大小数值符合你的loss计算公式 2. 随着模型训练, 整体loss是下降趋势并且模型测试的时候有结果出来 具体loss是个什么数值不是很关键, 满足上面两点就可以

@amitabhama 我tf跑了大概有50个epoch左右就没再继续训了, 跑去搞其它任务去了,不过之前训练的结果是实际测试起来效果挺好, 但ai val上score得分较差,后来没再跟了.

@amitabhama heatmap paf 定义在这[link](https://github.com/murdockhou/lightweight_openpose/blob/master/train.py#L70) loss肯定越小越好, 具体哪里会好不是很清楚了