qshi

confusiuon about get_affine_transform

2

https://github.com/microsoft/human-pose-estimation.pytorch/blob/18f1d0fa5b5db7fe08de640610f3fdbdbed8fb2f/lib/utils/transforms.py#L57 dst_dir = np.array([0, dst_w * -0.5], np.float32) dst[1, :] = np.array([dst_w * 0.5, dst_h * 0.5]) + dst_dir I can get `dst[1, 1] == (dst_h-dst_w)*0.5` but Why?? why not...


pytorch1.5 onnx2caffe failed

5

from pytorch1.5 to onnx,then convert onnx to caffemodel with your code but the output of caffemodel is zero why? [onnx_pytorchv1.5.zip](https://github.com/MTlab/onnx2caffe/files/4626459/onnx_pytorchv1.5.zip)

Running wrong with "main_prediction.py"

2

Hello，thank you for your share.but i encounter some problems as follows: No such file or directory: '/with SPN/conv1/W.npy' "self.load_weights('/Users/shashank/TensorFlow/with SPN/weights/')" "self.sess.run(var.assign(np.load(path+var.name.split('/')[0]+'/b.npy')))" I guess the lack of files！！！

do you know public dataset that can prove my algorithm works

 the mistake occurs when detecting nothing! The score of background is the highest!

Combine Bi-box_Regression and Repulsion loss

2

Do you try to combine two methods?

This is my result！

11

 这个结果和Repulsion loss一样都变差了，也是奇怪的很！希望作者贴出他的实验结果！