Research map / Detection and segmentation / Weakly supervised detection and segmentation
Diffusion-Assisted Progressive Learning for Weakly Supervised Phrase Localization
Pengyue Lin, Yanyang Hu, Xinjing Liu, Wenqi Jia, Fangxiang Feng, Ruifan Li
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 attention-driven loss LA (ablation) | DAPO-TAS · Flickr30K Entities | pointing game accuracy: 76.54 -> 78.11 | Improved |
| add diffusion-guided loss LD on top of LA (ablation) | DAPO-TAS · Flickr30K Entities | pointing game accuracy: 78.11 -> 79.63 | Improved |
| add task loss Ltask on top of LA and LD (ablation) | DAPO-TAS · Flickr30K Entities | pointing game accuracy: 78.11 -> 79.63 | Improved |
| train with diffusion-guided loss only, without Ltask (ablation) | DAPO-TAS · Flickr30K Entities | pointing game accuracy: 65.41 vs 79.63 | Hurt |
| add quantity-based scoring (QSM) alone (ablation) | DAPO-TAS · Flickr30K Entities | pointing game accuracy: 77.03 -> 77.25 | No clear effect |
| add semantic-based scoring (SSM) alone (ablation) | DAPO-TAS · Flickr30K Entities | pointing game accuracy: 77.03 -> 78.52 | Improved |
| combine both QSM and SSM (ablation) | DAPO-TAS · Flickr30K Entities | pointing game accuracy: 78.52 -> 79.63 | Improved |
| combine both QSM and SSM (ablation) | DAPO-TAS · VG | bounding box accuracy: 27.09 -> 32.08 | Improved |
| apply DAPO to VPT backbone | VPT · Flickr30K Entities | pointing game accuracy: 80.17 -> 85.43 | Improved |
| apply DAPO to WWbL-g on multi-object VG-M | WWbL-g · VG-M | pointing game accuracy: 42.05 -> 51.45 | Improved |
| apply DAPO to VPT on multi-object VG-M | VPT · VG-M | pointing game accuracy: 43.28 -> 54.50 | Improved |
| increase number of resampled phrases beyond 10^4 (ablation) | DAPO-TAS · Flickr30K Entities | accuracy | No clear effect |
Closest papers by meaning
- Align2Ground: Weakly Supervised Phrase Grounding Guided by Image-Caption Alignment (ICCV 2019 · similarity 0.90)
- Improving Weakly Supervised Visual Grounding by Contrastive Knowledge Distillation (CVPR 2021 · similarity 0.89)
- Boosting Weakly Supervised Referring Image Segmentation via Progressive Comprehension (NeurIPS 2024 · similarity 0.88)
- Siamese Learning with Joint Alignment and Regression for Weakly-Supervised Video Paragraph Grounding (CVPR 2024 · similarity 0.87)
- Momentum Pseudo-Labeling for Weakly Supervised Phrase Grounding (AAAI 2025 · similarity 0.86)
- Omnipotent Distillation with LLMs for Weakly-Supervised Natural Language Video Localization: When Divergence Meets Consistency (AAAI 2024 · similarity 0.86)
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.
