Research map / Video / Action recognition
HACS: Human Action Clips and Segments Dataset for Recognition and Temporal Localization
Hang Zhao, Antonio Torralba, Lorenzo Torresani, Zhicheng Yan
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 |
|---|---|---|---|
| pretrain on HACS Clips instead of Kinetics-600 (RGB) | I3D · Kinetics-400 val | Video@1 accuracy: 72.9 -> 73.4 | Improved |
| pretrain on HACS Clips instead of Kinetics-600 (RGB) | I3D · HMDB51 | Video@1 accuracy: 73.4 -> 73.6 | Improved |
| pretrain on HACS Clips instead of Kinetics-600 (Flow) | I3D · UCF101 | Video@1 accuracy: 96.0 -> 95.7 | Hurt |
| pretrain on HACS Clips instead of no pretraining (RGB) | I3D · HMDB51 | Video@1 accuracy: 39.4 -> 73.6 | Improved |
| pretrain on HACS Clips instead of no pretraining (RGB) | I3D · UCF101 | Video@1 accuracy: 75.0 -> 95.1 | Improved |
| add optical flow to RGB with late fusion (ablation) | I3D · HACS Clips val | Class@1: 80.3 -> 83.5 | Improved |
| sample clips with maximum consensus instead of random for training | I3D/Res3D · Validation sets | Class@1 | Improved |
| maximum consensus sampling instead of random sampling | clip sampling · Train-mini clips | number of positive clips: 82.2K -> 100.3K | Improved |
| maximum entropy sampling instead of random sampling | clip sampling · Train-mini clips | number of positive clips: 82.2K -> 71.3K | Hurt |
| train BSN on full HACS Segments instead of Mini (ablation) | BSN · HACS Segments | AR@100: 61.85 -> 63.62 | Improved |
| train SSN on full HACS Segments instead of Mini (ablation) | SSN · HACS Segments | average mAP: 15.93 -> 18.97 | Improved |
| evaluate on HACS Segments Mini instead of ActivityNet | BSN · localization difficulty | AR@100: 74.16 -> 61.85 | Hurt |
Closest papers by meaning
- Action Search: Spotting Actions in Videos and Its Application to Temporal Action Localization (ECCV 2018 · similarity 0.89)
- AVA: A Video Dataset of Spatio-Temporally Localized Atomic Visual Actions (CVPR 2018 · similarity 0.89)
- Learning to Localize Actions from Moments (ECCV 2020 · similarity 0.88)
- Event Stream based Human Action Recognition: A High-Definition Benchmark Dataset and Algorithms (IJCV 2026 · similarity 0.88)
- SCSampler: Sampling Salient Clips From Video for Efficient Action Recognition (ICCV 2019 · similarity 0.88)
- Graph Convolutional Module for Temporal Action Localization in Videos (TPAMI 2021 · 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.
