SkateFormer

SkateFormer copied to clipboard

Hi there, When I train the model from scratch, I can only achieve 93.45 on CS NTU-60, whereas the paper reports 93.6, which is a decline of 0.15. The only...

xiaobaoxia

Question about Partition Implementation (P2, P3, P4) in Code

1

Thank you for sharing your open-source code! While studying the partitioning strategy described in your paper, I noticed that the code implementation may not fully align with the description for...

JackWangPulse

Other modalities' pretrained weights

2

Hello there, I can see you have published the pre-trained weights only for Joint and Bone modalities. Would it be possible for you to also publish the pre-trained weights for...

soroushoraki

Hello @JeonghyeokDo 🤗 I'm Niels and I work on computer vision at Hugging Face. I see your paper is accepted as ECCV Poster, congratulations! I have indexed it here https://huggingface.co/papers/2403.09508...

NielsRogge

Train a custom dataset using SkateFormer

1

Hi dear all, is there any available script able to train a new dataset using your model pretrained or training a new model on new dataset from scratch? What would...

gguzzy

Could you specify torchlight version?

1

Hi, I would like to work with your project but am having problems with the environment. I've installed `torchlight`, using pip as per the requirements section, and have version 0.0.1...

shangong

Thank you guys for this work. Do you have an implementation of how to calculate the Skate-Type Importance Score ?

Madou-AI