seekFire

Question about the parameter β

1

Thank U for your excellent work! I have read your paper and modified some code in this repo so as to train models using my own dataset. You know the...

Issue about the dataset used in your code

5

In _crowd.py_, the code shows that the .npy is the annotation file format, so which public dataset does it belong to? I found the annotation files of many public dataset...

用tf2onnx将模型从h5转为onnx后出了点问题

2

使用tf2onnx API: tf2onnx.convert.from_keras() 将模型转为onnx格式后，发现模型输出的尺寸是输入的1/2，而不是1/4 用Netron查看onnx格式的模型，发现结构里有4次strides=2的卷积(3个depthwise + 1个常规)，3次ConvTranspose，但是配置文件里input_shape和output_shape却是按4倍降采样的设置......

How to create 3D custom dataset?

11

I noticed the python script(tools/data/custom_2d_skeleton.py) for creating 2D custom data, so how to create 3D dataset? or what script tools should we use?

输入GCN的tensor要求的人数是2个，但大多数的NTU动作是单人的，因此是需要把第一个人的keypoint信息复制到tensor中的第二个人的位置上。但此处的代码: pyskl/datasets/pipelines/pose_related.py:  第450行， 此处已经将单人的keypoint tensor和zero tensor进行了concatenate操作，得到的tensor已经是代表两个人的了。那下一行的判断条件里的keypoint.shape[0] 就不可能等于1，这还怎么执行随后的复制操作？ 还是说单人的动作，只需要给第二个人的位置填充zero tensor，不需要复制第一个人的信息？

Thank you for your work! I'm very interested in it. In your paper, the step 4 in algorithm of loss function is shown as follows: 4: compute L CDIoU =...