Research map / Detection and segmentation / 3D detection and driving
Contrastive Lift: 3D Object Instance Segmentation by Slow-Fast Contrastive Fusion
Yash Sanjay Bhalgat, Iro Laina, Joao F. Henriques, Andrea Vedaldi, Andrew Zisserman
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add slow-fast (EMA teacher) scheme to vanilla contrastive loss (ablation) | Contrastive Lift · Messy Rooms (avg) | PQ scene: 63.2 -> 66.5 | Improved |
| add concentration loss to slow-fast loss (ablation) | Contrastive Lift · Messy Rooms (avg) | PQ scene: 66.5 -> 69.0 | Improved |
| add concentration loss to slow-fast loss (ablation) | Contrastive Lift · ScanNet (avg) | PQ scene: 61.3 -> 62.0 | Improved |
| add slow-fast scheme to vanilla contrastive loss (ablation) | Contrastive Lift · ScanNet (avg) | PQ scene: 60.5 -> 61.3 | Improved |
| add concentration loss to vanilla contrastive without slow field (ablation) | Contrastive Lift vanilla · Messy Rooms (avg) | PQ scene: 63.2 -> 51.7 | Hurt |
| add concentration loss to vanilla contrastive without slow field (ablation) | Contrastive Lift vanilla · ScanNet (avg) | PQ scene: 60.5 -> 55.2 | Hurt |
| replace Panoptic Lifting linear assignment with slow-fast contrastive lifting | PanopLi · Messy Rooms Old Room 500 objects | PQ scene: 51.0 -> 55.0 | Improved |
| replace linear assignment with slow-fast contrastive lifting | PanopLi · ScanNet | PQ scene: 58.9 -> 62.3 | Improved |
| vanilla contrastive loss instead of linear assignment | PanopLi · Replica | PQ scene: 57.9 -> 57.8 | No clear effect |
| lift 2D Mask2Former segments to 3D with Contrastive Lift | Mask2Former · ScanNet | PQ (per frame): 42.0 -> 61.6 | Improved |
| replace HDBSCAN with DBSCAN clustering (ablation) | Contrastive Lift · ScanNet | PQ scene: 62.0 -> 61.8 | No clear effect |
| replace contrastive loss with DINO-style loss (ablation) | Contrastive Lift · ScanNet | PQ scene: 60.5 -> 54.7 | Hurt |
Closest papers by meaning
- Open3DIS: Open-Vocabulary 3D Instance Segmentation with 2D Mask Guidance (CVPR 2024 · similarity 0.87)
- Rethinking End-to-End 2D to 3D Scene Segmentation in Gaussian Splatting (CVPR 2025 · similarity 0.87)
- SAM2Object: Consolidating View Consistency via SAM2 for Zero-Shot 3D Instance Segmentation (CVPR 2025 · similarity 0.87)
- Insightful Instance Features for 3D Instance Segmentation (CVPR 2025 · similarity 0.86)
- Any3DIS: Class-Agnostic 3D Instance Segmentation by 2D Mask Tracking (CVPR 2025 · similarity 0.86)
- SAI3D: Segment Any Instance in 3D Scenes (CVPR 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
