search

huang-yh / GaussianFormer

[ECCV 2024] Scene as Gaussians for Vision-Based 3D Semantic Occupancy Prediction
Other
296 stars 22 forks source link

Self-encoding #4

Open AlanLiangC opened 5 months ago

AlanLiangC commented 5 months ago

Hello. Thank you for your excellent work. I would like to know how you handle gaussians that fall on the same grid during sparse convolution.

LoickCh commented 5 months ago

Hey, Not 100% sure about this but I think most of the traditional librairies compute the mean of features falling in the same voxel.

AlanLiangC commented 5 months ago

Hey, Not 100% sure about this but I think most of the traditional librairies compute the mean of features falling in the same voxel.

Hello, If this is the case, different positions of Gaussians may result in the same features, or how are the sparse features obtained from self-encoding utilized? I feel that the article does not make this clear. Could you help me?

huang-yh commented 2 months ago

Sorry for the late reply. We used implementation from spconv. Please check model/encoder/gaussian_encoder/spconv3d_module.py for more details.