search

open-mmlab / mmsegmentation

OpenMMLab Semantic Segmentation Toolbox and Benchmark.
https://mmsegmentation.readthedocs.io/en/main/
Apache License 2.0
8.34k stars 2.63k forks source link

How to change the code to make it support NYUDv2 dataset. #1178

Closed WenjieZhao1 closed 2 years ago

WenjieZhao1 commented 2 years ago

NYUDv2 dataset is a dataset which contains RGB images and depth images. I want use mmseg to deal with multimodal fusion question.

MengzhangLI commented 2 years ago

Currently mmseg only supports 1-channel or 3-channel input, thus 4 channel (GRB + depth) input is not supported yet.

But we have noticed this requirements from community, here is related issue: https://github.com/open-mmlab/mmsegmentation/issues/1121.

We plan to add this feature as soon as possible.

WenjieZhao1 commented 2 years ago

The common way for multimodal fusion is copy the depth channel into 3-channel, and then do the fusion operation with rgb-channel. So, the final input is 3-channel.