HOU Yuenan

You can just google the problem. It seems that the problem lies in the dependent packages.

Can you refer to the issues in that repo? It's hard for me to give you the answer as many old codes have been cleaned several days ago.

Sorry about the mistake. You can just randomly split one subset (around 10 K images) from the original training set as the validation set. The performance in our paper is...

Yes. The coefficient is selected to make sure that each transformed id is unique.

I am currently engaged in other projects and do not have spare time to clean the distillation codes. I will inform you when they are ready. You can have a...

You can first have a look at [this repo](https://github.com/HobbitLong/RepDistiller) to incorporate KD into the current code. The KD parts will be released when they are ready.

Thanks for your interest. We are still cleaning the codes since we want to release a plug-and-play toolbox. You can first have a try of the vanilla version to test...

You need to modify the feature_dim in the [self.PPmodel](https://github.com/cardwing/Codes-for-PVKD/blob/main/network/cylinder_fea_generator.py#L19), e.g., 64 to 32, 128 to 64, 256 to 128.

We just follow the practice of Cylinder3D.

In extractors.py, you can find the pre-trained weights in the following URLs: model_urls = { 'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth', 'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth', 'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth', 'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth', 'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth', } Hope this issue can...