Research map / Detection and segmentation / CNN features
Mask-Attention-Free Transformer for 3D Instance Segmentation
Xin Lai, Yuhui Yuan, Ruihang Chu, Yukang Chen, Han Hu, Jiaya Jia
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 iterative refinement of position queries (ablation) | Mask-attention-free transformer · ScanNetv2 val | mAP: 57.5 -> 58.4 | Improved |
| add center matching to bipartite matching (ablation) | Mask-attention-free transformer · ScanNetv2 val | mAP: 56.7 -> 58.4 | Improved |
| add center loss (ablation) | Mask-attention-free transformer · ScanNetv2 val | mAP: 56.8 -> 58.4 | Improved |
| add both center matching and center loss together with iterative refinement (ablation) | Mask-attention-free transformer · ScanNetv2 val | mAP: 56.4 -> 58.4 | Improved |
| use relative position encoding instead of Fourier absolute encoding (ablation) | Mask-attention-free transformer · ScanNetv2 val | mAP: 57.7 -> 58.4 | Improved |
| add any position encoding (versus none) (ablation) | Mask-attention-free transformer · ScanNetv2 val | mAP: 0.0 -> 58.4 | Improved |
| use learnable position queries instead of FPS-initialized (ablation) | Mask-attention-free transformer · ScanNetv2 val | mAP: 57.3 -> 58.4 | Improved |
| use zero-initialized instead of learnable content queries (ablation) | Mask-attention-free transformer · ScanNetv2 val | mAP: 58.1 -> 58.4 | Improved |
| replace mask attention with center regression and position-aware designs | Mask3D/SPFormer-style baseline · ScanNetv2 val | convergence: 128 epochs outperform baseline trained 512 epochs | Improved |
| add surface normal input | Mask-attention-free transformer · ScanNetv2 val | mAP: 58.4 -> 59.9 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- Mask3D — Mask3D: Pre-Training 2D Vision Transformers by Learning Masked 3D Priors (CVPR 2023) · better on 5 of 5 shared results
- CAGroup3D — CAGroup3D: Class-Aware Grouping for 3D Object Detection on Point Clouds (NeurIPS 2022) · 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
- Any3DIS: Class-Agnostic 3D Instance Segmentation by 2D Mask Tracking (CVPR 2025 · similarity 0.89)
- Query Refinement Transformer for 3D Instance Segmentation (ICCV 2023 · similarity 0.89)
- GTA: A Geometry-Aware Attention Mechanism for Multi-View Transformers (ICLR 2024 · similarity 0.87)
- Open3DIS: Open-Vocabulary 3D Instance Segmentation with 2D Mask Guidance (CVPR 2024 · similarity 0.87)
- Insightful Instance Features for 3D Instance Segmentation (CVPR 2025 · similarity 0.86)
- Segment Any 3D Object with Language (ICLR 2025 · similarity 0.86)
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.
