Pointcept / PointTransformerV3

[CVPR'24 Oral] Official repository of Point Transformer V3 (PTv3)
MIT License
583 stars 30 forks source link

SemanticKITTI config #38

Closed FT-Zheng closed 2 months ago

FT-Zheng commented 2 months ago

Thank you for your work! I tried to reproduce the config used by ptv3 on SemanticKITTI based on the ptv3-nuScenes and other SemanticKITTI configs, but the mIoU on SemanticKITTI val only reached 67.0. May I ask the channels and patch_size you set up on SemanticKITTI? I'm not sure what's wrong with the config I reproduce.

Gofinge commented 2 months ago

Hi, I have been quite busy recently, and I also haven't collected and reproduced the SemanticKITTI best performance config from my experimental code. Maybe you can try to remove PointClip from data augmentation. I think I will have some time one month later and release the last piece of our PTv3 task lists.

FT-Zheng commented 2 months ago

Thank you for your reply. I will try again.