Research map / Point clouds / Segmentation and detection
3D Instances as 1D Kernels
Yizheng Wu, Min Shi, Shuaiyuan Du, Hao Lu, Zhiguo Cao, Weicai Zhong
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 LN-NMS instead of random candidates (ablation) | DKNet · ScanNetV2 val | AP: 48.0 -> 50.8 | Improved |
| use LN-NMS instead of NMS (ablation) | DKNet · ScanNetV2 val | AP@50: 64.6 -> 66.7 | Improved |
| add aggregation loss during training only (ablation) | DKNet · ScanNetV2 val | AP@50: 62.6 -> 64.5 | Improved |
| aggregate duplicate candidates at inference too (ablation) | DKNet · ScanNetV2 val | AP@50: 64.5 -> 66.7 | Improved |
| aggregate candidate features instead of eliminating duplicates (ablation) | DKNet · ScanNetV2 val | AP@50: 65.2 -> 66.7 | Improved |
| use features from all points of instance instead of aggregated candidates (ablation) | DKNet · ScanNetV2 val | AP@50: 66.7 -> 67.0 | No clear effect |
| add coordinate information to kernel (ablation) | DKNet · ScanNetV2 val | mAP: 49.5 -> 50.8 | Improved |
| add shape information (foreground/background pooling) to kernel (ablation) | DKNet · ScanNetV2 val | mAP: 49.3 -> 50.8 | Improved |
| vary kernel shape/size from [8,1] to [16,16,1] (ablation) | DKNet · ScanNetV2 val | mAP: 50.6 -> 50.9 | No clear effect |
| add candidate aggregation (ablation) | DKNet · ScanNetV2 val detection | AP@50: 57.9 -> 59.0 | Improved |
| use CUDA LN-NMS instead of python | DKNet · inference time | ms: 187 ms -> 79 ms | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- DyCo3D — DyCo3D: Robust Instance Segmentation of 3D Point Clouds Through Dynamic Convolution (CVPR 2021) · better on 1 of 1 shared result
- 3D-MPA — 3D-MPA: Multi-Proposal Aggregation for 3D Semantic Instance Segmentation (CVPR 2020) · better on 1 of 1 shared result
- PointGroup — PointGroup: Dual-Set Point Grouping for 3D Instance Segmentation (CVPR 2020) · better on 1 of 1 shared result
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
- Open3DIS: Open-Vocabulary 3D Instance Segmentation with 2D Mask Guidance (CVPR 2024 · similarity 0.90)
- Insightful Instance Features for 3D Instance Segmentation (CVPR 2025 · similarity 0.89)
- Any3DIS: Class-Agnostic 3D Instance Segmentation by 2D Mask Tracking (CVPR 2025 · similarity 0.89)
- Spherical Mask: Coarse-to-Fine 3D Point Cloud Instance Segmentation with Spherical Representation (CVPR 2024 · similarity 0.89)
- Segment Any 3D Object with Language (ICLR 2025 · similarity 0.87)
- MaskClustering: View Consensus based Mask Graph Clustering for Open-Vocabulary 3D Instance Segmentation (CVPR 2024 · similarity 0.87)
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.
