Research map / Detection and segmentation / Weakly supervised detection and segmentation
Concept Mask: Large-Scale Segmentation from Semantic Concepts
Yufei Wang, Zhe Lin, Xiaohui Shen, Jianming Zhang, Scott Cohen
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 |
|---|---|---|---|
| finetune embedding network with refinement network (Original-Finetuned vs Original) (ablation) | attention map generation · OIVG-750 eval set | Pointing Game / IoU: 0.578 -> 0.631 (Pointing Game), 0.262 -> 0.288 (IoU) | Improved |
| add the two-layer refinement network trained on box annotations (Refined-noNeg vs Original-Finetuned) (ablation) | attention map generation · OIVG-750 eval set | Pointing Game / IoU: 0.631 -> 0.806 (Pointing Game), 0.288 -> 0.416 (IoU) | Improved |
| add negative examples from non-overlapping concepts (spatial discrimination loss) (ablation) | refinement network · OIVG-750 eval set | Pointing Game / IoU: 0.806 -> 0.810 (Pointing Game), 0.416 -> 0.421 (IoU) | Improved |
| add negative examples from non-overlapping concepts and multi-scale attention input (full model vs Ours-noNeg and Ours-singleAtt) (ablation) | segmentation model · COCO-80, Weak-Box-670, Weak-Image-50 | IoU | Improved |
| use concept-conditioned attention model instead of saliency detection (DSS) on unseen concepts | Concept Mask · ZeroTest-10 | IoU: 0.298 -> 0.436 | Improved |
| reduce supervision level from full masks to boxes to image-level labels | Concept Mask · COCO-80 / Weak-Box-670 / Weak-Image-50 | segmentation IoU | Hurt |
| use label-agnostic Mask-RCNN* instead of attention-conditioned two-stage model | weakly supervised segmentation · Weak-Box-670 | IoU | Hurt |
Closest papers by meaning
- Single-Stage Semantic Segmentation From Image Labels (CVPR 2020 · similarity 0.90)
- Open-World Semantic Segmentation via Contrasting and Clustering Vision-Language Embedding (ECCV 2022 · similarity 0.90)
- Webly Supervised Semantic Segmentation (CVPR 2017 · similarity 0.90)
- Weakly Supervised Semantic Segmentation via Box-Driven Masking and Filling Rate Shifting (TPAMI 2023 · similarity 0.89)
- Reliability Does Matter: An End-to-End Weakly Supervised Semantic Segmentation Approach (AAAI 2020 · similarity 0.89)
- Semantic-aware SAM for Point-Prompted 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.
