Research map / Video / Video-language understanding
Text-Driven Video Acceleration: A Weakly-Supervised Reinforcement Learning Method
Washington Ramos, Michel Silva, Edson Araujo, Victor Moura, Keller Oliveira, Leandro Soriano Marcolino, Erickson R. Nascimento
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 SAFFA instead of BoT | BoT · YouCook2 | F1 score: 14.04 -> 17.86 | Improved |
| use SAFFA instead of SAS | SAS · COIN | F1 score: 13.20 -> 17.18 | Improved |
| use SAFFA instead of dense-supervised FFNet | FFNet · YouCook2 | F1 score: 18.86 -> 17.86 | No clear effect |
| use SAFFA instead of FFNet | FFNet · YouCook2 | Recall: 11.70 -> 12.81 | Improved |
| use SAFFA instead of BoT (running time) | BoT · YouCook2/COIN | running time (ms): 1048.01 -> 75.58 | Improved |
| add skip-aware vector to the agent state (ablation) | SAFFA with VDAN+ · YouCook2 | Overall Performance: 00.00 -> 26.33 | Improved |
| add NRPE on top of skip-aware state (ablation) | SAFFA with VDAN+ · YouCook2 | Overall Performance: 26.33 -> 30.07 | Improved |
| add NRPE on top of skip-aware state (ablation) | SAFFA with VDAN+ · YouCook2 | F1 score: 15.18 -> 17.86 | Improved |
| use VDAN+ instead of VDAN semantic encoder (ablation) | SAFFA (SA+NRPE) · YouCook2 | F1 score: 14.33 -> 17.86 | Improved |
| use VDAN+ instead of VDAN for frame/clip-document classification (ablation) | VDAN · YouCook2 train | AUC: 61 -> 65 | Improved |
| initialize h0 with phi(v) instead of zeros (ablation) | VDAN+ semantic encoder · VaTeX validation | MRR: 0.065 -> 0.718 | Improved |
| initialize h0 with phi(v) instead of learnable W_s (ablation) | VDAN+ semantic encoder · VaTeX validation | MRR: 0.061 -> 0.718 | Improved |
Closest papers by meaning
- Straight to the Point: Fast-Forwarding Videos via Reinforcement Learning Using Textual Data (CVPR 2020 · similarity 0.95)
- Read, Watch, and Move: Reinforcement Learning for Temporally Grounding Natural Language Descriptions in Videos (AAAI 2019 · similarity 0.87)
- Efficient Frame Selection for Long Video Understanding via Reinforcement Learning (CVPR 2026 · similarity 0.87)
- Weakly Supervised Video Representation Learning With Unaligned Text for Sequential Videos (CVPR 2023 · similarity 0.87)
- VSTAR: Generative Temporal Nursing for Longer Dynamic Video Synthesis (ICLR 2025 · similarity 0.86)
- Video-STaR: Self-Training Enables Video Instruction Tuning with Any Supervision (ICLR 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.
