Research map / Video / Action recognition
Listen to Look: Action Recognition by Previewing Audio
Ruohan Gao, Tae-Hyun Oh, Kristen Grauman, Lorenzo Torresani
Check 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 iterative attention-based moment selection instead of dense prediction aggregation | ImgAud-Skimming · ActivityNet | accuracy (%): 66.3 -> 70.3 | Improved |
| Use iterative attention-based moment selection instead of dense prediction aggregation | ImgAud-Skimming · Mini-Sports1M | accuracy (%): 37.3 -> 39.2 | Improved |
| Use context-aware moment selection instead of selecting the top confidence clips | ImgAud-Skimming · ActivityNet | accuracy (%): 69.1 -> 70.3 | Improved |
| Use context-aware moment selection instead of selecting the top confidence clips | ImgAud-Skimming · Mini-Sports1M | accuracy (%): 38.4 -> 39.2 | Improved |
| Replace recurrent iterative queries with a single query (ablation) | ImgAud-Skimming · ActivityNet | accuracy (%): 70.3 -> 67.5 | Hurt |
| Replace recurrent iterative queries with a single query (ablation) | ImgAud-Skimming · Mini-Sports1M | accuracy (%): 39.2 -> 38.0 | Hurt |
| Process all image-audio features sequentially rather than selecting useful moments (ablation) | ImgAud-Skimming · ActivityNet | accuracy (%): 70.3 -> 63.5 | Hurt |
| Process all image-audio features sequentially rather than selecting useful moments (ablation) | ImgAud-Skimming · Mini-Sports1M | accuracy (%): 39.2 -> 34.1 | Hurt |
| Use image-audio indexing with R(2+1)D-152 recognition instead of dense evaluation | ImgAud-Skimming · ActivityNet | mAP: 88.9 -> 88.5 | Hurt |
| Use image-audio indexing with R(2+1)D-152 recognition instead of dense evaluation | ImgAud-Skimming · ActivityNet | TFLOPs: 25.9 -> 1.31 | Improved |
| Use R(2+1)D-152 features for both indexing and recognition instead of dense evaluation | ImgAud-Skimming · ActivityNet | mAP: 88.9 -> 89.9 | Improved |
| Use R(2+1)D-152 features for both indexing and recognition instead of dense evaluation | ImgAud-Skimming · ActivityNet | TFLOPs: 25.9 -> 2.64 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SCSampler — SCSampler: Sampling Salient Clips From Video for Efficient Action Recognition (ICCV 2019) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SCSampler: Sampling Salient Clips From Video for Efficient Action Recognition (ICCV 2019 · similarity 0.88)
- ${\text{CA}^{2}\text{ST}}$: Cross-Attention in Audio, Space, and Time for Holistic Video Recognition (TPAMI 2025 · similarity 0.88)
- Cooperative Learning of Audio and Video Models from Self-Supervised Synchronization (NeurIPS 2018 · similarity 0.87)
- TIM: A Time Interval Machine for Audio-Visual Action Recognition (CVPR 2024 · similarity 0.87)
- Beyond Short Clips: End-to-End Video-Level Learning With Collaborative Memories (CVPR 2021 · similarity 0.87)
- Action Recognition From Single Timestamp Supervision in Untrimmed Videos (CVPR 2019 · 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.
