NYU Dataset with 4 channel depth images.

zifuwan / Sigma

[WACV 2025] Python implementation of Sigma: Siamese Mamba Network for Multi-Modal Semantic Segmentation

https://zifuwan.github.io/Sigma/

MIT License

190 stars 19 forks source link

NYU Dataset with 4 channel depth images. #23

Closed yuwenmichael closed 4 months ago

yuwenmichael commented 4 months ago

Hi,

I downloaded your preprocessed dataset from google drive and found that the depth image is of 4 channels? May I know what exact channel(s) you pass to the model for training and testing?

Also I noted that in the config file https://github.com/zifuwan/Sigma/blob/fcacbd92d0f2c3f8c09d79c1e12631d577134a30/configs/config_nyu.py#L32, the model takes only one channel for training if it is correct, I was wondering which channel you referring to?

Thanks,

zifuwan commented 4 months ago

Hi, we read the depth image with grey mode (1 channel) and copy it to 3 channels before passing to the model. Please refer to this code: link Hope this can be useful.