Some questions about json files from raw data and inputs of the 3D U-Net

[weihaoSun1999]

Hello, authors. Your work is very exciting and meaningful! I have some questions about the details in the json file. How is point_coord selected? How are features of the volume, surface_area, radius gotten? Are they taken from the centerline file?
In addition, I have some questions about the inputs of the 3D U-Net. Since KP2 in one of the small isolated components can be found using morphological operations, we can randomly select a point in that small object as the center point of our subvolume. However, how can we locate 808080 subvolume of the disconnected point in KP1 part. Since the location of KP1 and KP2 components is unknown, how to develop a sampling strategy in the 3D U-Net.
I'm looking forward to your Pytorch implementation of keypoint detection. May I ask the release time? I believe your work will be of great significance for the task of pulmonary airways and vessels.
Best wishes!

[weihaoSun1999]

May I ask the release time of the pytorch implementation keypoint detection?
@duducheng @ZiqiaoWeng

[Zhongrocky]

@weihaoSun1999 Hello, Have you implemented this code yet？