hustvl / Symphonies

[CVPR 2024] Symphonies (Scene-from-Insts): Symphonize 3D Semantic Scene Completion with Contextual Instance Queries
https://arxiv.org/abs/2306.15670
MIT License
168 stars 6 forks source link

Adaptation of another dataset #20

Closed SuNy4 closed 6 months ago

SuNy4 commented 6 months ago

Hello, firstly thanks for new insights through your work. I have a question for adapting another dataset.

As far as I know, SSC voxel map of SemanticKITTI has a trail of moving object, providers called spatio-temporal tube. I think this may affect metric scores or even performances of your work. So I am trying to adapt Occ3D dataset to your work expecting better performances. (But It's not a stereo dataset) Have you tried this dataset before or planned to adapt in future?

Again, thanks for your work! lots of things to learn from it.😊

npurson commented 6 months ago

Hi there, thanks for your interest in our work!

Regarding Occ3D, we're indeed working on it for our future work, but regrettably, we don't have any plans to adapt Symphonies to it at present.

BTW, I believe that multi-views on nuScenes can utilize even more geometrical cues than stereo does. Besides, there are also some more advanced metric depth estimation methods, such as Metric3D V2.

Please let us know if you have any other questions or progress!

SuNy4 commented 6 months ago

Thanks for kind reply!

Your advice and opinion make me reconsider about geometrical enrichment. I'll let you know if we had any progress!

Also I am looking forward your future work! Thanks again!😊