Hongyu Pan

Sorry. The reason why I reimplemented the approach is for some other tasks. So, I didn't evaluate on the COCO Dataset.

Based on the error message, I think the problem is not relevant to the version. Obviously, you have gotten the keys, which are just different from the keys of the...

Maybe you can try change mask = np.ones((1, 1, input_img.shape[2] / stride, input_img.shape[3] / stride), dtype=np.float32) from mask = np.ones((1, 1, int(input_img.shape[2] / stride), int(input_img.shape[3] / stride)), dtype=np.float32) But I'm...

Don't understand the question. And I think the cause of the error is the data error. Please check the data and debug.

Sorry, I don't know. This is about PyTorch, and I don't know how to save the middle parameters with PyTorch, like caffe's solverstate or tf's checkpoint. So I don't know...

If you just want to use the middle model without middle state information, you can load the middle model directly to initialize the model.

change the line 49 "model = torch.nn.DataParallel(model, device_ids=args.gpu).cuda()" in train_pose.py to "model = model.cuda()"

Yes, you can.

I don't know why there are so many blue (purple?) dots. The purple dot means right ear. You could try if don't draw the eye and ear. I don't think...

make sure the test image has substracted mean and been divided by std. What's more, pay attention to the padValue.