Research map / Detection and segmentation / Open-set and few-shot detection
Active Coarse-to-Fine Segmentation of Moveable Parts from Real Images
Ruiqi Wang, Akshay Gadi Patil, Fenggen Yu, Hao Zhang
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 object mask head to transformer baseline (ablation) | OPDFormer-C-style baseline · Our dataset test set | segm mAP: 68.4 -> 74.9 | Improved |
| add pose and interaction direction heads (ablation) | Ours w/o AL · Our dataset test set | segm mAP: 74.9 -> 77.3 | Improved |
| add human-in-the-loop active learning (ablation) | Ours w/o AL · Our dataset test set | segm mAP: 77.3 -> 91.3 | Improved |
| use coarse-to-fine AL instead of fine-only AL (ablation) | Ours · 500-image enhancement set | total lab time (hr): 1.675 -> 1.6 | Improved |
| use coarse-to-fine AL instead of fine-only AL (ablation) | Ours · 2,000-image set | total lab time (hr): 7.5 -> 6.5 | Improved |
| use coarse-to-fine AL instead of fine-only AL (ablation) | Ours · Our dataset test set | segm mAP: 91.2 -> 91.3 | No clear effect |
| add object mask head on top of pose/direction heads with coarse-to-fine AL (ablation) | Ours · Our dataset test set | total lab time (hr): 2.4 -> 1.6 | Improved |
| add active learning to OPDFormer-C | OPDFormer-C · 500-image enhancement set | total lab time (hr): 5.62 -> 3.9 | Improved |
| use proposed network instead of OPDFormer-C without AL | OPDFormer-C · OPDReal test | segm mAP: 46.3 -> 51.6 | Improved |
| use zero-shot Grounded-SAM without fine-tuning on part data | Grounded-SAM · Our dataset test set | segm mAP: 23.1 vs 45.2 (OPD-C) | Hurt |
Closest papers by meaning
- Real-Time Monocular Segmentation and Pose Tracking of Multiple Objects (ECCV 2016 · similarity 0.83)
- Physical Regularization Loss: Integrating Physical Knowledge to Image Segmentation (IJCV 2026 · similarity 0.82)
- Coarse-To-Fine Segmentation With Shape-Tailored Continuum Scale Spaces (CVPR 2017 · similarity 0.82)
- Unsupervised Co-part Segmentation through Assembly (ICML 2021 · similarity 0.81)
- MOVE: Unsupervised Movable Object Segmentation and Detection (NeurIPS 2022 · similarity 0.81)
- VOSTR: Video Object Segmentation via Transferable Representations (IJCV 2020 · similarity 0.81)
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.
