Research map / Detection and segmentation / Weakly supervised detection and segmentation
DBGroup: Dual-Branch Point Grouping for Weakly Supervised 3D Semantic Instance Segmentation
Xuexun Liu, Xiaoxu Xu, Qiudan Zhang, Lin Ma, Xu Wang
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 |
|---|---|---|---|
| use GAIM instead of Bidirectional Merging (ablation) | DBGroup · ScanNetV2 val (pseudo-label ablation, T=1) | AP: 16.3 -> 17.4 | Improved |
| add MGB to SGB with GAIM (ablation) | DBGroup · ScanNetV2 val (pseudo-label ablation, T=1) | AP: 16.3 -> 17.4 | Improved |
| add SGB to MGB with GAIM (ablation) | DBGroup · ScanNetV2 val (pseudo-label ablation, T=1) | AP: 15.5 -> 17.4 | Improved |
| select top-alpha points class-wise (ablation) | DBGroup · ScanNetV2 val (SSP/IMF ablation, T=1) | AP: 23.6 -> 25.0 | Improved |
| add superpoint propagation (ablation) | DBGroup · ScanNetV2 val (SSP/IMF ablation, T=1) | AP: 25.0 -> 26.5 | Improved |
| add Instance Mask Filter (ablation) | DBGroup · ScanNetV2 val (SSP/IMF ablation, T=1) | AP: 25.4 -> 26.5 | Improved |
| increase self-training iterations from 1 to 3 (ablation) | DBGroup · ScanNetV2 val (self-training iterations) | AP: 26.5 -> 28.6 | Improved |
| increase self-training iterations from 3 to 5 (ablation) | DBGroup · ScanNetV2 val (self-training iterations) | AP: 28.6 -> 28.8 | No clear effect |
| set GAIM overlap threshold 0.4 instead of 0.2 (ablation) | DBGroup · ScanNetV2 val (pseudo-label ablation, T=1) theta | AP: 17.1 -> 17.4 | Improved |
| set GAIM overlap threshold 0.4 instead of 0.6 (ablation) | DBGroup · ScanNetV2 val (pseudo-label ablation, T=1) theta | AP: 16.7 -> 17.4 | Improved |
| use BFS grouping instead of class-wise DBSCAN (ablation) | DBGroup · ScanNetV2 val (pseudo-label ablation, T=1) | AP: 17.1 -> 17.4 | Improved |
| keep top 30% instead of 10% in semantic selection (ablation) | DBGroup · ScanNetV2 val (SSP alpha ablation) | AP: 21.4 -> 26.5 | Improved |
Closest papers by meaning
- Instance Consistency Regularization for Semi-Supervised 3D Instance Segmentation (TPAMI 2024 · similarity 0.91)
- Sketchy Bounding-box Supervision for 3D Instance Segmentation (CVPR 2025 · similarity 0.89)
- P2Seg: Pointly-supervised Segmentation via Mutual Distillation (ICLR 2024 · similarity 0.89)
- Insightful Instance Features for 3D Instance Segmentation (CVPR 2025 · similarity 0.89)
- Segment Any 3D Object with Language (ICLR 2025 · similarity 0.88)
- Semantic-aware SAM for Point-Prompted Instance Segmentation (CVPR 2024 · similarity 0.88)
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.
