3D Semantic Segmentation on Indoor datasets joint (train)
71.5mIoUPoint-MoE-L
Evaluation Results
| Method | Links | |
|---|---|---|
| Point-MoE-LFLOPs/step (GFLOPs)=265.7, Peak VRAM (GiB)=33.3, ∆ FLOPs vs. PPT-L=↓ 30.9%, ∆ VRAM vs. PPT-L=↓ 19.0%2025.05 | 71.5 |
Loading the SOTA2 catalog…