Research map / Detection and segmentation / Weakly supervised detection and segmentation
Box2Mask: Box-Supervised Instance Segmentation via Level-Set Evolution
Wentong Li, Wenyu Liu, Jianke Zhu, Miaomiao Cui, Ri‐Sheng Yu, Xian‐Sheng Hua, Lei 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 image data term to box-projection-only level-set (ablation) | Box2Mask-C · Pascal VOC val | mask AP: 27.1 -> 30.6 | Improved |
| add deep feature data term on top of image term (ablation) | Box2Mask-C · Pascal VOC val | mask AP: 30.6 -> 33.8 | Improved |
| restrict level-set evolution to the bounding box instead of full image (image term) (ablation) | Box2Mask-C · Pascal VOC val | mask AP: 28.3 -> 30.6 | Improved |
| restrict level-set evolution to the bounding box instead of full image (image+feature terms) (ablation) | Box2Mask-C · Pascal VOC val | mask AP: 28.8 -> 33.8 | Improved |
| tree filter guided by low-level image only (ablation) | Box2Mask-C · Pascal VOC val | mask AP: 30.8 -> 28.2 | Hurt |
| tree filter guided by low and high-level features (ablation) | Box2Mask-C · Pascal VOC val | mask AP: 30.8 -> 33.8 | Improved |
| add local consistency module with dilation 3 (ablation) | Box2Mask-C · Pascal VOC val | mask AP: 33.8 -> 36.3 | Improved |
| enlarge LCM dilation rate to 5 (ablation) | Box2Mask-C · Pascal VOC val | mask AP: 36.3 -> 33.5 | Hurt |
| longer 3x multi-scale schedule (ablation) | Box2Mask-C · Pascal VOC val | mask AP: 36.3 -> 38.0 | Improved |
| longer 50-epoch schedule (ablation) | Box2Mask-T · Pascal VOC val | mask AP: 39.4 -> 41.4 | Improved |
| increase MSDeformAttn layers from 1 to 2 (ablation) | Box2Mask-T · Pascal VOC val | mask AP: 36.8 -> 38.9 | Improved |
| increase MSDeformAttn layers from 5 to 6 (ablation) | Box2Mask-T · Pascal VOC val | mask AP: 39.4 -> 39.3 | No clear effect |
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.
- BoxInst — BoxInst: High-Performance Instance Segmentation With Box Annotations (CVPR 2021) · better on 6 of 6 shared results
- DiscoBox — DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence From Box Supervision (ICCV 2021) · better on 5 of 5 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
- Box-Supervised Instance Segmentation with Level Set Evolution (ECCV 2022 · similarity 0.96)
- BoxInst: High-Performance Instance Segmentation With Box Annotations (CVPR 2021 · similarity 0.93)
- BoxTeacher: Exploring High-Quality Pseudo Labels for Weakly Supervised Instance Segmentation (CVPR 2023 · similarity 0.91)
- SIM: Semantic-Aware Instance Mask Generation for Box-Supervised Instance Segmentation (CVPR 2023 · similarity 0.91)
- Learning to Segment Every Thing (CVPR 2018 · similarity 0.89)
- Extreme Point Supervised Instance 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.
