Research map / Video / Action recognition
MaCLR: Motion-Aware Contrastive Learning of Representations for Videos
Fanyi Xiao, Joseph Tighe, Davide Modolo
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 |
|---|---|---|---|
| add Motion pathway and motion-visual contrastive loss (ablation) | Visual-only contrastive · UCF101 (K400-mini) | top-1 accuracy: 63.6 -> 78.1 | Improved |
| add Motion pathway and motion-visual contrastive loss (ablation) | Visual-only contrastive · HMDB51 (K400-mini) | top-1 accuracy: 33.7 -> 47.2 | Improved |
| train with full K400 instead of K400-mini plus motion (ablation) | Visual-only contrastive · UCF101 (K400) | top-1 accuracy: 74.6 -> 85.5 | Improved |
| add Motion pathway (ablation) | Visual-only contrastive · HMDB51 (K400) | top-1 accuracy: 46.3 -> 57.7 | Improved |
| use flow edge maps instead of frame difference as motion input (ablation) | MaCLR · UCF101 | top-1 accuracy: 71.6 -> 78.1 | Improved |
| use flow edge maps instead of raw flow as motion input (ablation) | MaCLR · UCF101 | top-1 accuracy: 74.1 -> 78.1 | Improved |
| use flow edge maps instead of raw flow (ablation) | MaCLR · HMDB51 | top-1 accuracy: 44.2 -> 47.2 | Improved |
| add temporal jittering (ablation) | MaCLR · UCF101 | top-1 accuracy: 77.4 -> 78.1 | Improved |
| add temporal jittering (ablation) | MaCLR · HMDB51 | top-1 accuracy: 47.1 -> 47.2 | No clear effect |
| add motion threshold sampling (ablation) | MaCLR · UCF101 | top-1 accuracy: 77.3 -> 78.1 | Improved |
| add motion threshold sampling (ablation) | MaCLR · HMDB51 | top-1 accuracy: 46.4 -> 47.2 | Improved |
| add motion-only loss L_m (ablation) | MaCLR · UCF101 | top-1 accuracy: 77.8 -> 78.1 | Improved |
Closest papers by meaning
- Motion-Focused Contrastive Learning of Video Representations (ICCV 2021 · similarity 0.90)
- Frame-Wise Action Representations for Long Videos via Sequence Contrastive Learning (CVPR 2022 · similarity 0.89)
- Self-Supervised Video Representation Learning by Context and Motion Decoupling (CVPR 2021 · similarity 0.88)
- Contrast and Order Representations for Video Self-Supervised Learning (ICCV 2021 · similarity 0.88)
- Cross-Architecture Self-Supervised Video Representation Learning (CVPR 2022 · similarity 0.88)
- Self-Supervised Video Representation Learning in a Heuristic Decoupled Perspective (IJCV 2026 · 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.
