Research map / Point clouds / Segmentation and detection
3D Instance Segmentation via Enhanced Spatial and Semantic Supervision
Salwa Al Khatib, Mohamed El Amine Boudjoghra, Jean Lahoud, Fahad Shahbaz Khan
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 spatial and semantic supervision plus Vr in mask prediction to Mask3D | Mask3D · ScanNet200 val | mAP50: 37.0 -> 39.7 | Improved |
| same changes on ScanNet V2 | Mask3D · ScanNet V2 val | mAP50: 73.7 -> 75.0 | Improved |
| same changes, tail classes | Mask3D · ScanNet200 tail | mAP50: 23.2 -> 29.1 | Improved |
| same changes, head classes | Mask3D · ScanNet200 head | mAP50: 54.9 -> 55.5 | Improved |
| append voxel coordinates Vr in mask prediction (ablation) | spatial-supervision model · ScanNet200 val | mAP50: 37.1 -> 39.6 | Improved |
| add spatial supervision (ablation) | mask prediction with Vr · ScanNet200 val | mAP50: 39.2 -> 39.6 | Improved |
| add semantic supervision (ablation) | spatial + Vr · ScanNet200 val | mAP50: 39.6 -> 39.7 | Improved |
| add spatial supervision to semantic + Vr (ablation) | semantic + Vr · ScanNet200 val | mAP50: 38.4 -> 39.7 | Improved |
| use spatial supervision instead of coordinates as input (ablation) | backbone input · ScanNet200 val | mAP50: 36.2 -> 38.0 | Improved |
| add coordinates to color input vs coordinates alone (ablation) | backbone input · ScanNet200 val | mAP50: 36.2 -> 36.6 | Improved |
| add all three components | Mask3D · ScanNet200 val | mAP: 27.4 -> 29.4 | Improved |
Closest papers by meaning
- Dynamic Convolution for 3D Point Cloud Instance Segmentation (TPAMI 2022 · similarity 0.92)
- DyCo3D: Robust Instance Segmentation of 3D Point Clouds Through Dynamic Convolution (CVPR 2021 · similarity 0.92)
- OneFormer3D: One Transformer for Unified Point Cloud Segmentation (CVPR 2024 · similarity 0.91)
- PointInst3D: Segmenting 3D Instances by Points (ECCV 2022 · similarity 0.91)
- Insightful Instance Features for 3D Instance Segmentation (CVPR 2025 · similarity 0.90)
- Segment Any 3D Object with Language (ICLR 2025 · similarity 0.89)
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.
