Research map / Video / Video-language understanding
Co-Grounding Networks With Semantic Attention for Referring Expression Comprehension in Videos
Sijie Song, Xudong Lin, Jiaying Liu, Zongming Guo, Shih-Fu Chang
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 baseline grounding with learned subject attention (ablation) | One-stage LSTM baseline · VID | Acc.@0.5: 54.41 -> 58.03 | Improved |
| add learned subject and location attention (ablation) | One-stage LSTM baseline · VID | Acc.@0.5: 54.41 -> 59.22 | Improved |
| use an external semantic parser instead of learned semantic attention (ablation) | One-stage LSTM baseline · VID | Acc.@0.5: 54.41 -> 53.19 | Hurt |
| add co-grounding feature learning (ablation) | One-stage LSTM baseline · VID | Acc.@0.5: 54.41 -> 55.88 | Improved |
| add co-grounding feature learning to subject attention (ablation) | S-Att. · VID | Acc.@0.5: 58.03 -> 58.74 | Improved |
| add co-grounding feature learning to subject and location attention (ablation) | SL-Att. · LiOTB | Acc.@0.5: 50.51 -> 50.92 | Improved |
| add post-processing to co-grounding with semantic attention (ablation) | CG-SL-Att. · VID | Acc.@0.5: 59.48 -> 60.25 | Improved |
| increase the post-processing window from P=1 to P=5 (ablation) | CG-SL-Att. · VID | Acc.@0.5: 59.48 -> 60.25 | Improved |
| increase the post-processing window from P=5 to P=9 (ablation) | CG-SL-Att. · VID | Acc.@0.5: 60.25 -> 59.63 | Hurt |
| apply co-grounding with the proposed framework rather than one-stage LSTM | Ours · VID | Accu.@0.5: 54.78 -> 60.25 | Improved |
| apply co-grounding with the proposed framework rather than one-stage LSTM | Ours · LiOTB | Accu.@0.5: 49.16 -> 52.26 | Improved |
| add semantic attention to the Darknet53 baseline (ablation) | SL-Att. · RefCOCO val | Acc.@0.5: 73.72 -> 77.65 | Improved |
Closest papers by meaning
- Referring Transformer: A One-step Approach to Multi-task Visual Grounding (NeurIPS 2021 · similarity 0.89)
- Comprehensive Visual Grounding for Video Description (AAAI 2024 · similarity 0.88)
- Referencing Where to Focus: Improving Visual Grounding with Referential Query (NeurIPS 2024 · similarity 0.88)
- Multi-Pair Temporal Sentence Grounding via Multi-Thread Knowledge Transfer Network (AAAI 2025 · similarity 0.88)
- Empower Words: DualGround for Structured Phrase and Sentence-Level Temporal Grounding (NeurIPS 2025 · similarity 0.88)
- RefDetector: A Simple Yet Effective Matching-based Method for Referring Expression Comprehension (AAAI 2025 · 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.
