search

omniobject3d / OmniObject3D

[ CVPR 2023 Award Candidate ] OmniObject3D: Large-Vocabulary 3D Object Dataset for Realistic Perception, Reconstruction and Generation
462 stars 12 forks source link

The number of categories #31

Closed Yebulabula closed 10 months ago

Yebulabula commented 10 months ago

Hi,

Thanks for publishing this very fantastic dataset.

I noticed a discrepancy in the dataset's category count. While there are 216 subfolders in the HDF5 point cloud folder, suggesting 216 categories, the filenames within these subfolders don't consistently match the subfolder names. Counting unique filenames indicates only 145 categories. Could you clarify which category count should be used for point cloud classification tasks using your dataset?

Thanks.