[Suggestion] Benchmarking Humanoid Perception with OccStudio

[cdb342]

Hi team, fantastic work on the Humanoid Occupancy dataset! 🚀

I wanted to recommend OccStudio (https://github.com/cdb342/OccStudio). We thought our framework's high configurability might be helpful for your team or the community working with your dataset.

Why OccStudio?

• 🔧 Flexible & Configurable Architecture: We support diverse configurations switchable via config files, including input modalities (Images, Depth), 3D encodings (Volume-based, BEV-based), temporal fusion (SoloFusion, GDFusion), and backbones (ResNet, Swin, InterImage).
• 🏆 Unified SOTA Models: A single codebase for ALOcc (ICCV'25), GDFusion (CVPR'25), CausalOcc (ICCV'25), BEVDet, FB-Occ, etc.

We hope this codebase can serve as a convenient platform for running SOTA benchmarks on your impressive dataset!