Research map / Point clouds / Segmentation and detection
ISBNet: A 3D Point Cloud Instance Segmentation Network With Instance-Aware Sampling and Box-Aware Dynamic Convolution
Tuan Duc Ngo, Binh-Son Hua, Khoi Nguyen
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 ISBNet instead of the strongest baseline | ScanNetV2 hidden test · ScanNetV2 hidden test | AP: 55.9 vs. 53.2 | Improved |
| Use ISBNet instead of the strongest baseline | S3DIS 6-fold cross-validation · S3DIS 6-fold cross-validation | AP: 60.8 vs. 54.4 | Improved |
| Use ISBNet instead of the strongest baseline | STPLS3D validation · STPLS3D validation | AP: 49.2 vs. 46.2 | Improved |
| Replace standard FPS with IA-FPS (ablation) | ISBNet baseline on ScanNetV2 validation · ScanNetV2 validation | AP: 47.9 -> 49.7 | Improved |
| Add Local Aggregation Layer (ablation) | ISBNet baseline on ScanNetV2 validation · ScanNetV2 validation | AP: 47.9 -> 50.1 | Improved |
| Replace standard dynamic convolution with box-aware dynamic convolution (ablation) | ISBNet baseline on ScanNetV2 validation · ScanNetV2 validation | AP: 47.9 -> 48.6 | Improved |
| Combine IA-FPS and Local Aggregation Layer (ablation) | ISBNet baseline on ScanNetV2 validation · ScanNetV2 validation | AP: +5.5 AP; +6.8 AP50 | Improved |
| Add box regression auxiliary supervision (ablation) | ISBNet on ScanNetV2 validation · ScanNetV2 validation | AP: 52.8 -> 53.4 | Improved |
| Use predicted boxes as an extra geometric cue in dynamic convolution (ablation) | ISBNet on ScanNetV2 validation · ScanNetV2 validation | AP: 53.4 -> 54.5 | Improved |
| Use three Point Aggregator blocks instead of two (ablation) | ISBNet on ScanNetV2 validation · ScanNetV2 validation | AP: 54.5 -> 54.3 | Hurt |
| Use only one dynamic convolution layer (ablation) | ISBNet on ScanNetV2 validation · ScanNetV2 validation | AP: 54.5 -> 45.7 | Hurt |
| Sample 512 candidates in one IA-FPS chunk (ablation) | ISBNet on ScanNetV2 validation · ScanNetV2 validation | AP: 54.2 -> 53.6 | Hurt |
Later papers that compared with this method
1 later accepted paper reports ISBNet as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Spherical Mask: Coarse-to-Fine 3D Point Cloud Instance Segmentation with Spherical Representation (CVPR 2024 · better on 2 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- FreePoint: Unsupervised Point Cloud Instance Segmentation (CVPR 2024 · similarity 0.90)
- Open3DIS: Open-Vocabulary 3D Instance Segmentation with 2D Mask Guidance (CVPR 2024 · similarity 0.89)
- Insightful Instance Features for 3D Instance Segmentation (CVPR 2025 · similarity 0.89)
- Relation3D : Enhancing Relation Modeling for Point Cloud Instance Segmentation (CVPR 2025 · similarity 0.88)
- LinNet: Linear Network for Efficient Point Cloud Representation Learning (NeurIPS 2024 · similarity 0.88)
- OneFormer3D: One Transformer for Unified Point Cloud 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.
