Research map / Video / Video-language understanding
Temporal Modular Networks for Retrieving Complex Compositional Activities in Videos
Bingbin Liu, Serena Yeung, Edward Chou, De-An Huang, Li Fei-Fei, Juan Carlos Niebles
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 TMN modular compositional network instead of MCN | MCN · DiDeMo RGB | Rank@5: 44.82 -> 72.97 | Improved |
| use TMN instead of MCN | MCN · DiDeMo RGB | Rank@1: 13.10 -> 18.71 | Improved |
| use TMN instead of MCN | MCN · DiDeMo Flow | Rank@1: 18.35 -> 19.90 | Improved |
| use TMN instead of MCN | MCN · DiDeMo Fuse | Rank@1: 19.88 -> 22.92 | Improved |
| use TMN instead of MCN | MCN · DiDeMo Fuse | mean IoU: 33.51 -> 35.17 | Improved |
| add combine attention instead of max pooling (ablation) | TMN const + rank loss · DiDeMo intra-video (Fuse) | Rank@1: 21.89 -> 22.92 | Improved |
| add combine attention instead of max pooling (ablation) | TMN const + rank loss · DiDeMo intra-video (Fuse) | Rank@5: 75.69 -> 76.08 | Improved |
| use ranking loss instead of BCE loss (ablation) | TMN const + attention · DiDeMo intra-video (Fuse) | Rank@1: 21.60 -> 22.92 | Improved |
| use ranking loss instead of BCE loss (ablation) | TMN dep + attention · DiDeMo intra-video (Fuse) | Rank@1: 20.41 -> 21.67 | Improved |
| use constituency parser instead of dependency parser (ablation) | TMN attention + rank loss · DiDeMo intra-video (Fuse) | Rank@1: 21.67 -> 22.92 | Improved |
| use constituency parser instead of dependency parser (ablation) | TMN attention + rank loss · DiDeMo intra-video (Fuse) | mean IoU: 33.94 -> 35.17 | Improved |
| add compositional tree structure (vs MCN, no tree) (ablation) | TMN · DiDeMo intra-video (Fuse) | Rank@5: 62.39 -> 75.69 | Improved |
Closest papers by meaning
- VELOCITI: Benchmarking Video-Language Compositional Reasoning with Strict Entailment (CVPR 2025 · similarity 0.88)
- Compositional Video Understanding with Spatiotemporal Structure-based Transformers (CVPR 2024 · similarity 0.87)
- Vgent: Graph-based Retrieval-Reasoning-Augmented Generation For Long Video Understanding (NeurIPS 2025 · similarity 0.86)
- STEP: Enhancing Video-LLMs' Compositional Reasoning by Spatio-Temporal Graph-guided Self-Training (CVPR 2025 · similarity 0.86)
- HENASY: Learning to Assemble Scene-Entities for Interpretable Egocentric Video-Language Model (NeurIPS 2024 · similarity 0.86)
- From Play to Replay: Composed Video Retrieval for Temporally Fine-Grained Videos (NeurIPS 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.
