Hongyu Pan

Another difference about PAF. paper says:  the code is:  If don't update the value of cnt when cnt = 0. The value of cnt always will be 0....

The mask has been generated. You can generate the file by youself. And you can remove this line.

It needs to be generated by yourself.

@wangfei-cs please see README

I have added the code for generating masklist/train2017.txt. Please see the [script](https://github.com/last-one/Pytorch_Realtime_Multi-Person_Pose_Estimation/blob/master/preprocessing/generate_json_mask.py)

Because the evaluation of my task doesn't use mAP, I didn't write the script. You can see my testing script, which could get the estimated points and edges. Hope it...

If you use one gpu to train and change line 49 "model = torch.nn.DataParallel(model, device_ids=args.gpu).cuda()" to "model = model.cuda()", you're right. Because of torch.nn.DataParallel, it will change "model" to "model.module"...

No, I think. If you find something error, please let me know. Thank you.

Yes. The test and train scripts should be changed. I have mentioned something in README.

The shape of feature (x) and the gt (y), which will be used to calculate the loss, is N * 46 * 46 (if input size is 368, 46 =...