Research map / Video / Action recognition
Compressed Video Action Recognition
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Train on compressed-video streams (I-frames, motion vectors, and residuals) instead of RGB alone (ablation) | CoViAR · UCF-101 | accuracy: 87.7 -> 90.4 | Helped |
| Train on compressed-video streams (I-frames, motion vectors, and residuals) instead of RGB alone (ablation) | CoViAR · HMDB-51 | accuracy: 53.3 -> 59.1 | Helped |
| Add motion vectors to I-frame RGB (ablation) | CoViAR · UCF-101 | accuracy: 87.7 -> 89.9 | Helped |
| Add residuals to I-frame RGB (ablation) | CoViAR · UCF-101 | accuracy: 87.7 -> 89.7 | Helped |
| Add residuals to I-frame RGB (ablation) | CoViAR · HMDB-51 | accuracy: 53.3 -> 55.2 | Helped |
| Add motion vectors and residuals to I-frame RGB (ablation) | CoViAR · UCF-101 | accuracy: +2.7 | Helped |
| Add motion vectors and residuals to I-frame RGB (ablation) | CoViAR · HMDB-51 | accuracy: +5.8 | Helped |
| Use accumulated rather than original motion vectors (ablation) | CoViAR motion stream · UCF-101 split 1 | accuracy: 58.3 -> 63.9 | Helped |
| Use accumulated rather than original residuals (ablation) | CoViAR residual stream · UCF-101 split 1 | accuracy: 79.0 -> 79.9 | Helped |
| Use accumulated motion vectors and residuals (ablation) | Full CoViAR · UCF-101 split 1 | accuracy: 90.4 -> 90.8 | Helped |
| Replace decoupled processing with Conv-LSTM aggregation (ablation) | CoViAR · UCF-101 split 1 | accuracy: 90.8 -> 89.1 | Hurt |
| Add an input-to-output skip connection to Conv-LSTM (ablation) | Conv-LSTM variant · UCF-101 split 1 | accuracy: 89.1 -> 87.8 | Hurt |
Closest papers by meaning
- DMC-Net: Generating Discriminative Motion Cues for Fast Compressed Video Action Recognition (CVPR 2019 · similarity 0.89)
- More Is Less: Learning Efficient Video Representations by Big-Little Network and Depthwise Temporal Aggregation (NeurIPS 2019 · similarity 0.89)
- Self-Supervised Learning of Compressed Video Representations (ICLR 2021 · similarity 0.88)
- No Frame Left Behind: Full Video Action Recognition (CVPR 2021 · similarity 0.87)
- Video Compression With Entropy-Constrained Neural Representations (CVPR 2023 · similarity 0.87)
- Action Recognition in Video Using Sparse Coding and Relative Features (CVPR 2016 · similarity 0.87)
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.