search

Pointcept / PointTransformerV2

[NeurIPS'22] An official PyTorch implementation of PTv2.
357 stars 25 forks source link

Dose all 10000 points of every objects in Modelnet40_normal_resampled dataset are used as the input? #13

Closed LiJin73 closed 1 year ago

LiJin73 commented 1 year ago

Thank you for your great job. But I have seen the code and the config in Issue #9, there seems to be no operation to limit the number of points for the Modelnet dataset.

Gofinge commented 1 year ago

Yes, our code focuses more on real-world scene-level point clouds and we just simply test our performance on ModelNet40 without too much tuning. You can add a point number limitation in transform.py or modelnet40.py, which might lead to better performance since 10000 points are too redundant for ModelNet40.