Research map / Video / Action recognition
Time-Equivariant Contrastive Video Representation Learning
Simon Jenni, Hailin Jin
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 temporal equivariance (temporal only) instead of none (ablation) | contrastive video model R3D-18 · UCF101 linear | accuracy: 54.2 -> 74.1 | Improved |
| add temporal equivariance (temporal only) instead of none (ablation) | contrastive video model R3D-18 · HMDB51 linear | accuracy: 25.2 -> 47.5 | Improved |
| add spatial equivariance instead of none (ablation) | contrastive video model R3D-18 · UCF101 linear | accuracy: 54.2 -> 62.8 | Improved |
| use temporal equivariance only instead of spatial + temporal (ablation) | contrastive video model R3D-18 · UCF101 linear | accuracy: 70.7 -> 74.1 | Improved |
| add auxiliary SSL losses to equivariance loss (ablation) | L_equi · UCF101 linear | accuracy: 36.2 -> 67.1 | Improved |
| add equivariance loss to instance + auxiliary losses (ablation) | L_inst + L_aux · UCF101 linear | accuracy: 71.8 -> 74.1 | Improved |
| add equivariance loss to instance + auxiliary losses (ablation) | L_inst + L_aux · UCF101 finetune | accuracy: 83.0 -> 83.7 | Mixed |
| add instance discrimination loss to equivariance + auxiliary losses (ablation) | L_equi + L_aux · UCF101 linear | accuracy: 67.1 -> 74.1 | Improved |
| add auxiliary losses to instance discrimination (ablation) | L_inst · UCF101 linear | accuracy: 54.2 -> 71.8 | Improved |
| add clip order/overlap task to speed + reverse auxiliary (ablation) | aux tasks · UCF101 linear | accuracy: 71.1 -> 74.1 | Improved |
| add reverse task to speed + order auxiliary (ablation) | aux tasks · UCF101 linear | accuracy: 70.1 -> 74.1 | Improved |
| add order task to speed auxiliary (ablation) | aux tasks · UCF101 linear | accuracy: 69.0 -> 70.1 | Improved |
Closest papers by meaning
- Contrast and Order Representations for Video Self-Supervised Learning (ICCV 2021 · similarity 0.90)
- Spatiotemporal Contrastive Video Representation Learning (CVPR 2021 · similarity 0.90)
- Self-Supervised Video Representation Learning via Latent Time Navigation (AAAI 2023 · similarity 0.90)
- Probabilistic Representations for Video Contrastive Learning (CVPR 2022 · similarity 0.89)
- Self-Supervised Video Representation Learning in a Heuristic Decoupled Perspective (IJCV 2026 · similarity 0.89)
- Frame-Wise Action Representations for Long Videos via Sequence Contrastive Learning (CVPR 2022 · similarity 0.89)
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.
