Research map / Vision-language and multimodal / Human and embodied modeling
Query-Guided Regression Network With Context Policy for Phrase Grounding
Kan Chen, Rama Kovvuri, Ram Nevatia
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 |
|---|---|---|---|
| replace retrieval GroundeR with query-guided regression (QRN) on RPN proposals (ablation) | RPN proposals, VGG-det · Flickr30K Entities | accuracy (%): 39.13 -> 53.48 | Improved |
| replace retrieval GroundeR with QRN on Selective Search proposals (ablation) | SS proposals, VGG-det · Flickr30K Entities | accuracy (%): 47.81 -> 55.99 | Improved |
| use jointly trained proposal generation network (PGN) instead of Selective Search (ablation) | QRN · Flickr30K Entities | accuracy (%): 55.99 -> 60.21 | Improved |
| add context policy network (CPN) with reinforcement-learning rewards (ablation) | PGN+QRN · Flickr30K Entities | accuracy (%): 60.21 -> 65.14 | Improved |
| add QRN regression and CPN on EdgeBoxes/PGN proposals | QRC Net · Referit Game | accuracy (%): 26.93 -> 44.07 | Improved |
| use PGN instead of EdgeBoxes for QRN (ablation) | QRN · Referit Game | accuracy (%): 32.21 -> 43.57 | Improved |
| add CPN (ablation) | PGN+QRN · Referit Game | accuracy (%): 43.57 -> 44.07 | Improved |
| use PGN proposals instead of RPN proposals (ablation) | proposal generation · Flickr30K Entities | UBP (%): 71.25 -> 89.61 | Improved |
| increase regression loss weight lambda from 1.0 to 10.0 (ablation) | QRC Net · Flickr30K Entities | accuracy (%): 65.14 -> 63.27 | Hurt |
| increase multimodal dimension m from 512 to 1024 (ablation) | QRC Net · Flickr30K Entities | accuracy (%): 65.14 -> 62.52 | Hurt |
| increase CPN reward beta from 0.2 to 0.8 (ablation) | QRC Net · Flickr30K Entities | accuracy (%): 65.14 -> 62.77 | Hurt |
| vary lambda from 0.5 to 4.0 (ablation) | QRC Net · Flickr30K Entities | accuracy (%): 64.15 / 65.14 / 64.40 / 64.29 | No clear effect |
Closest papers by meaning
- Iterative Shrinking for Referring Expression Grounding Using Deep Reinforcement Learning (CVPR 2021 · similarity 0.88)
- Knowledge Aided Consistency for Weakly Supervised Phrase Grounding (CVPR 2018 · similarity 0.88)
- Grounding of Textual Phrases in Images by Reconstruction (ECCV 2016 · similarity 0.87)
- Adaptive Reconstruction Network for Weakly Supervised Referring Expression Grounding (ICCV 2019 · similarity 0.87)
- Improving Zero-Shot Phrase Grounding via Reasoning on External Knowledge and Spatial Relations (AAAI 2022 · similarity 0.86)
- Diversifying Query: Region-Guided Transformer for Temporal Sentence Grounding (AAAI 2025 · 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.
