Research map / Detection and segmentation / CNN features
Segment and Recognize Anything at Any Granularity
Feng Li, Hao Zhang, Peize Sun, Xueyan Zou, Shilong Liu, Chunyuan Li, Jianwei Yang, Lei Zhang, Jianfeng Gao
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 SA-1B interactive segmentation data (1/10) to COCO panoptic joint training (ablation) | Semantic-SAM (T) · COCO val2017 instance | box AP: 50.1 -> 52.3 | Improved |
| add SA-1B data to COCO panoptic joint training (ablation) | Semantic-SAM (T) · COCO val2017 panoptic | PQ: 54.6 -> 55.2 | Improved |
| add SA-1B data to Pascal Part training (ablation) | Semantic-SAM (T) · Pascal Part | box AP: 27.0 -> 28.0 | Improved |
| use many-to-many matching instead of many-to-one for multi-granularity clicks (ablation) | Semantic-SAM (T) · SA-1B 1000-image subset | 1-IoU@All Granularity: 73.2 -> 88.1 | Improved |
| increase SA-1B training portion from 3% to 100% (ablation) | Semantic-SAM (L) · COCO val2017 point (Oracle) | 1-IoU: 73.5 -> 74.2 | No clear effect |
| train on SA-1B only with 6 granularity queries per click vs SAM with 3 masks | Semantic-SAM (L) · COCO val2017 point (Oracle) | 1-IoU: 70.5 -> 74.2 | Improved |
| Semantic-SAM (T) vs VLPart (B) box AP on Pascal Part | Semantic-SAM (T) · Pascal Part | box AP: 29.2 -> 27.0 | Hurt |
Closest papers by meaning
- SAM-CP: Marrying SAM with Composable Prompts for Versatile Segmentation (ICLR 2025 · similarity 0.89)
- Learning Hierarchical Image Segmentation For Recognition and By Recognition (ICLR 2024 · similarity 0.89)
- Semantic-aware SAM for Point-Prompted Instance Segmentation (CVPR 2024 · similarity 0.88)
- Segment Anything without Supervision (NeurIPS 2024 · similarity 0.88)
- GraCo: Granularity-Controllable Interactive Segmentation (CVPR 2024 · similarity 0.88)
- Bridge the Points: Graph-based Few-shot Segment Anything Semantically (NeurIPS 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.
