About train and infer in indoor dataset?

NVlabs / VoxFormer

Official PyTorch implementation of VoxFormer [CVPR 2023 Highlight]

Other

1.04k stars 86 forks source link

About train and infer in indoor dataset? #16

Closed someone-TJ closed 1 year ago

someone-TJ commented 1 year ago

How Can I train this model in NYUv2? I find the data format is so different from semantickitti

RoboticsYimingLi commented 1 year ago

You may refer to the dataloader in MonoScene (https://github.com/astra-vision/MonoScene/blob/master/monoscene/data/NYU/nyu_dataset.py) for NYUv2.

someone-TJ commented 1 year ago

Thanks for the suggestion, but this method seems to not get the information about queries, voxels, etc. needed for the training phase. I think there should be a general solution to the problem of voxformer running on different SSC datasets, as you mentioned https://github.com/NVlabs/VoxFormer/issues/7#issuecomment-1538191897, when will you guys release this toolkit?