Research map / Point clouds / Segmentation and detection
DyCo3D: Robust Instance Segmentation of 3D Point Clouds Through Dynamic Convolution
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace PointGroup's bottom-up instances with proposal-free dynamic convolution | DyCo3D Ours-S vs PointGroup · ScanNetV2 validation | mAP: 34.8 -> 35.4 | Helped |
| Use the large DyCo3D model instead of 3D-MPA | ScanNetV2 validation · ScanNetV2 validation | mAP: 35.3 -> 40.6 | Helped |
| Use DyCo3D instead of PointGroup | ScanNetV2 test · ScanNetV2 test | AP@50: 0.636 -> 0.641 | Helped |
| Use DyCo3D instead of PointGroup | S3DIS Area-5 · S3DIS Area-5 | mPrec: 61.9 -> 64.3 | Helped |
| Use DyCo3D instead of PointGroup | S3DIS Area-5 · S3DIS Area-5 | mRec: 62.1 -> 64.2 | Helped |
| Use dynamic filters generated from grouped homogeneous points instead of point-wise filters without clustering (ablation) | DyCo3D · ScanNetV2 validation | mAP: 24.8 -> 29.4 | Helped |
| Add instance-related position embeddings (ablation) | CondInst-style baseline · ScanNetV2 validation | mAP: 24.8 -> 27.0 | Helped |
| Add position embeddings to grouped-point decoding (ablation) | DyCo3D · ScanNetV2 validation | mAP: 29.4 -> 31.8 | Helped |
| Add category-aware decoding (ablation) | DyCo3D · ScanNetV2 validation | mAP: 31.8 -> 34.1 | Helped |
| Add a lightweight transformer to the backbone (ablation) | DyCo3D · ScanNetV2 validation | mAP: 34.1 -> 35.4 | Helped |
| Reduce PointGroup's clustering radius from 3cm to 2cm | PointGroup · ScanNetV2 | mAP: -6.3% | Hurt |
| Reduce PointGroup's clustering radius from 3cm to 1cm | PointGroup · ScanNetV2 | mAP: -23.9% | Hurt |
Closest papers by meaning
- FreePoint: Unsupervised Point Cloud Instance Segmentation (CVPR 2024 · similarity 0.91)
- OneFormer3D: One Transformer for Unified Point Cloud Segmentation (CVPR 2024 · similarity 0.90)
- Spherical Mask: Coarse-to-Fine 3D Point Cloud Instance Segmentation with Spherical Representation (CVPR 2024 · similarity 0.90)
- Open3DIS: Open-Vocabulary 3D Instance Segmentation with 2D Mask Guidance (CVPR 2024 · similarity 0.89)
- Relation3D : Enhancing Relation Modeling for Point Cloud Instance Segmentation (CVPR 2025 · similarity 0.89)
- Insightful Instance Features for 3D Instance Segmentation (CVPR 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.