Research map / Detection and segmentation / Weakly supervised detection and segmentation
Beyond Semantic to Instance Segmentation: Weakly-Supervised Instance Segmentation via Semantic Knowledge Transfer and Self-Refinement
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 PAM to produce instance cues instead of conventional CAM cues (ablation) | BESTIE · PASCAL VOC 2012 val | mAP50: 12.9 -> 29.3 | Helped |
| Add instance-aware guidance so center and offset losses apply only to labeled instances (ablation) | BESTIE with PAM · PASCAL VOC 2012 val | mAP50: 29.3 -> 39.2 | Helped |
| Add online self-refinement to IAG training (ablation) | BESTIE with PAM and IAG · PASCAL VOC 2012 val | mAP50: 39.2 -> 41.5 | Helped |
| Add center clustering when generating refined labels (ablation) | BESTIE with PAM, IAG, and self-refinement · PASCAL VOC 2012 val | mAP50: 41.5 -> 41.8 | Helped |
| Replace PMM semantic segmentation with SingleStage (ablation) | BESTIE · PASCAL VOC 2012 val | mAP50: 41.8 -> 39.7 | Hurt |
| Replace PMM semantic segmentation with RRM (ablation) | BESTIE · PASCAL VOC 2012 val | mAP50: 41.8 -> 41.1 | Hurt |
| Replace PMM output with ground-truth semantic segmentation (ablation) | BESTIE · PASCAL VOC 2012 val | mAP50: 41.8 -> 49.4 | Helped |
| Add iterative retraining after online self-refinement (ablation) | BESTIE · PASCAL VOC 2012 val | mAP50: 41.8 -> 41.9 -> 41.9 | No clear effect |
| Retain self-refinement but remove IAG (ablation) | BESTIE · PASCAL VOC 2012 val | mAP50: 41.8 -> 27.8 | Hurt |
| Use HRNet-48 instead of ResNet-50 (ablation) | BESTIE · PASCAL VOC 2012 val | mAP50: 40.9% -> 41.8% | Helped |
| Use PAM only in the last classifier layer rather than in the last three layers (ablation) | BESTIE · PASCAL VOC 2012 val | mAP50: 34.8% -> 41.8% | Helped |
| Vary PAM threshold alpha from 0.3 to 0.7 (ablation) | BESTIE · PASCAL VOC 2012 val | mAP50: 41.76; 41.80; 41.72 | No clear effect |
Closest papers by meaning
- WISH: Weakly Supervised Instance Segmentation using Heterogeneous Labels (CVPR 2025 · similarity 0.91)
- P2Seg: Pointly-supervised Segmentation via Mutual Distillation (ICLR 2024 · similarity 0.89)
- MWSIS: Multimodal Weakly Supervised Instance Segmentation with 2D Box Annotations for Autonomous Driving (AAAI 2024 · similarity 0.89)
- Semantic-aware SAM for Point-Prompted Instance Segmentation (CVPR 2024 · similarity 0.89)
- Progressive Feature Self-Reinforcement for Weakly Supervised Semantic Segmentation (AAAI 2024 · similarity 0.89)
- W2P: Switching from Weak Supervision to Partial Supervision for Semantic Segmentation (AAAI 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.