🎯 DiCoSeg Author : Jiawei Lian, Zhengxue Wang, Wentao Qu, Haobo Jiang, Le Hui, Jian Yang
🔬 Tackling 3D segmentation under annotation scarcity? DiCoSeg leverages diffusion models for context-aware reconstruction with just a few labels.
⚡️ Single-step inference, strong generalization, and robust results across benchmarks.
📈 Currently under submission — stay tuned for updates!
📬 If you're reviewing this... be kind 😄
📎 [Code coming soon]
| Settings | Method | mIoU | Where |
|---|---|---|---|
| 0.01% | PointNext | 58.4 | NeurIPS |
| SQN | 45.3 | ECCV | |
| CPCM | 59.3 | ICCV | |
| AADNet | 60.8 | AAAI | |
| DiCoSeg (Ours) | 61.1 | logs / ckpt |
| Settings | Method | mIoU | Where |
|---|---|---|---|
| 0.1% | SQN | 61.4 | ECCV |
| CPCM | 66.3 | ICCV | |
| PointCT | 68.3 | WACV | |
| AADNet | 67.2 | AAAI | |
| DiCoSeg (Ours) | 68.3 | logs / - |
| Settings | Method | mIoU | Where |
|---|---|---|---|
| 1% | RandLA-Net | 59.8 | CVPR |
| SQN | 63.6 | ECCV | |
| PointCT | 67.6 | WACV | |
| DiCoSeg (Ours) | 68.3 | logs / - |