chensuzeyu / SpotOccView on GitHub
To solve the decoder bottleneck in sparse 3D occupancy prediction, we propose SPOT-Occ. It uses a prototype-based sparse Transformer decoder, replacing costly dense attention with a "guided selection + focused aggregation" process. Queries focus on a small set of key "prototypes". The final model is significantly faster and more accurate.
17Feb 5, 2026Updated 5 months ago

Alternatives and similar repositories for SpotOcc

Users that are interested in SpotOcc are comparing it to the libraries listed below. We may earn a commission when you buy through links labeled 'Ad' on this page.

Sorting:

Are these results useful?