Research map / Detection and segmentation / Open-set and few-shot detection
Revisiting Image-Language Networks for Open-Ended Phrase Detection
Bryan A. Plummer, Kevin J. Shih, Yichen Li, Ke Xu, Svetlana Lazebnik, Stan Sclaroff, Kate Saenko
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 |
|---|---|---|---|
| initialize region/phrase layers with CCA (ablation) | SimNet · Flickr30K Entities (phrase detection) | mAP (mean/total): 5.9 -> 12.7 | Improved |
| initialize layers with CCA (ablation) | EmbNet · Flickr30K Entities (phrase detection) | mAP (mean/total): 5.4 -> 12.3 | Improved |
| initialize layers with CCA (ablation) | SimNet · ReferIt Game (phrase detection) | mAP (mean/total): 4.7 -> 7.4 | Improved |
| initialize layers with CCA (ablation) | SimNet · Visual Genome (phrase detection) | mAP (mean/total): 2.2 -> 3.1 | Improved |
| initialize layers with CCA (ablation) | SimNet · Flickr30K Entities (phrase localization) | accuracy (overall): 71.7 -> 71.9 | No clear effect |
| initialize layers with CCA (ablation) | SimNet · ReferIt Game (phrase localization) | accuracy (overall): 59.0 -> 57.8 | Hurt |
| initialize layers with CCA (ablation) | SimNet · Visual Genome (phrase localization) | accuracy (overall): 36.5 -> 35.4 | Hurt |
| add negative phrase augmentation (NPA) (ablation) | QA R-CNN · Flickr30K Entities (phrase detection) | mAP (mean/total): 5.7 -> 5.9 | No clear effect |
| add negative phrase augmentation (NPA) (ablation) | QA R-CNN · Flickr30K Entities (phrase localization) | accuracy (overall): 71.1 -> 69.7 | Hurt |
| use linear CCA instead of Deep CCA as region classifier (ablation) | CCA · Flickr30K Entities (phrase detection) | mAP (mean/total): 9.6 -> 12.1 | Improved |
| add inverse frequency sampling (IFS) instead of all phrases (ablation) | SimNet + CCA init · Flickr30K Entities (phrase detection, PPA) | mAP (mean/total): 13.7 -> 14.2 | Improved |
| add inverse frequency sampling (IFS) instead of random subsampling (RS) (ablation) | SimNet + CCA init · Flickr30K Entities (phrase detection, PPA) | mAP (mean/total): 13.6 -> 14.2 | Improved |
Closest papers by meaning
- Zero-Shot Grounding of Objects From Natural Language Queries (ICCV 2019 · similarity 0.92)
- Improving Zero-Shot Phrase Grounding via Reasoning on External Knowledge and Spatial Relations (AAAI 2022 · similarity 0.91)
- Phrase Localization Without Paired Training Examples (ICCV 2019 · similarity 0.91)
- Multi-Level Multimodal Common Semantic Space for Image-Phrase Grounding (CVPR 2019 · similarity 0.90)
- Improving Weakly Supervised Visual Grounding by Contrastive Knowledge Distillation (CVPR 2021 · similarity 0.90)
- Improved Visual Grounding through Self-Consistent Explanations (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.
