Research map / Detection and segmentation / Weakly supervised detection and segmentation
Align2Ground: Weakly Supervised Phrase Grounding Guided by Image-Caption Alignment
Samyak Datta, Karan Sikka, Anirban Roy, Karuna Ahuja, Devi Parikh, Ajay Divakaran
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 |
|---|---|---|---|
| use topk Local Matching instead of max (permInv aggregator) (ablation) | Align2Ground · COCO caption-to-image retrieval | R@1: +16.3 R@1 | Improved |
| use topk Local Matching instead of max (permInv aggregator) (ablation) | Align2Ground · Flickr30k caption-to-image retrieval | R@1: +20.6 R@1 | Improved |
| use topk Local Matching instead of attention or max (ablation) | Align2Ground · COCO and Flickr30k | C2I and phrase localization | Improved |
| use topk instead of max with the sequence encoder for matched RoIs (ablation) | Align2Ground · COCO and Flickr30k | C2I | Improved |
| use phrases instead of words in Pooling-based matching (ablation) | Pooling-based · COCO | Det. (%): +0.1 Det% | Improved |
| use phrases instead of words in Pooling-based matching (ablation) | Pooling-based · Flickr30k | Det. (%): +0.5 Det% | Improved |
| use phrases instead of words in Pooling-based matching (ablation) | Pooling-based · COCO caption-to-image retrieval | R@1 | Improved |
| use Pooling-based (phrases) instead of Global embedding | Global · COCO caption-to-image retrieval | R@1 (%): 39.3 -> 48.4 | Improved |
| use Pooling-based (phrases) instead of Global embedding | Global · COCO | Det. (%): 12.2 -> 10.8 | Hurt |
| use Pooling-based (phrases) instead of Global embedding | Global · Flickr30k | Det. (%): 8.0 -> 8.9 | Improved |
| use Align2Ground instead of Global for phrase localization | Global · COCO | Det. (%): +2.5 | Improved |
| use Align2Ground instead of Pooling-based (phrases) for phrase localization | Pooling-based (phrases) · COCO | Det. (%): +3.9 | Improved |
Closest papers by meaning
- Contrastive Learning for Weakly Supervised Phrase Grounding (ECCV 2020 · similarity 0.93)
- Weakly-Supervised Visual Grounding of Phrases With Linguistic Structures (CVPR 2017 · similarity 0.92)
- What is Where by Looking: Weakly-Supervised Open-World Phrase-Grounding without Text Inputs (NeurIPS 2022 · similarity 0.92)
- Momentum Pseudo-Labeling for Weakly Supervised Phrase Grounding (AAAI 2025 · similarity 0.90)
- Improved Visual Grounding through Self-Consistent Explanations (CVPR 2024 · similarity 0.89)
- Cycle-Consistency Learning for Captioning and Grounding (AAAI 2024 · similarity 0.89)
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.
