Research map / Detection and segmentation / Weakly supervised detection and segmentation
Phrase Localization Without Paired Training Examples
Josiah Wang, Lucia Specia
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 tfoid detector to tfcoco (ablation) | Detector combination · Flickr30kEntities | accuracy: 37.57 -> 48.20 | Improved |
| add places365 detector (ablation) | CC+OI (w2v-max, union) · Flickr30kEntities | accuracy: 48.20 -> 50.49 | Improved |
| add colour detector (ablation) | CC+OI+PL (avg, union) · Flickr30kEntities | accuracy: 50.27 -> 49.61 | No clear effect |
| use 80 COCO categories instead of 20 (ablation) | tfcoco · Flickr30kEntities | accuracy: 36.33 -> 37.57 | No clear effect |
| use union instead of largest box strategy (ablation) | CC+OI · Flickr30kEntities | accuracy: 30.32 -> 48.20 | Improved |
| use union instead of consensus strategy (ablation) | CC+OI+PL+CL w2v-avg · Flickr30kEntities | accuracy: 49.51 -> 49.61 | No clear effect |
| add colour detector (ablation) | CC+OI+PL (avg, union), colour-term phrases · Flickr30kEntities (colour-term subset) | accuracy: 43.43 -> 41.81 | Hurt |
| add colour detector to scene phrases (ablation) | CC+OI+PL (avg, union) · Flickr30kEntities (colour-term subset, scene) | accuracy: 44.09 -> 58.06 | Improved |
| add tfoid to tfcoco (ablation) | Detector combination · ReferItGame | accuracy: 15.40 -> 19.82 | Improved |
| add places365 (ablation) | CC+OI · ReferItGame | accuracy: 21.21 -> 23.95 | Improved |
| add colour detector (ablation) | CC+OI+PL (largest) · ReferItGame | accuracy: 23.95 -> 26.48 | Improved |
| use largest box instead of consensus (ablation) | CC+OI+PL+CL · ReferItGame | accuracy: 25.52 -> 26.48 | Improved |
Closest papers by meaning
- Phrase Localization and Visual Relationship Detection With Comprehensive Image-Language Cues (ICCV 2017 · similarity 0.92)
- Align2Ground: Weakly Supervised Phrase Grounding Guided by Image-Caption Alignment (ICCV 2019 · similarity 0.91)
- Revisiting Image-Language Networks for Open-Ended Phrase Detection (TPAMI 2020 · similarity 0.91)
- What is Where by Looking: Weakly-Supervised Open-World Phrase-Grounding without Text Inputs (NeurIPS 2022 · similarity 0.90)
- Improved Visual Grounding through Self-Consistent Explanations (CVPR 2024 · similarity 0.89)
- Boosting Weakly Supervised Referring Image Segmentation via Progressive Comprehension (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.
