Research map / Video / Video segmentation and tracking
Spatiotemporal Contrastive Video Representation Learning
Check 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 |
|---|---|---|---|
| Add both spatial and temporal augmentation instead of using temporal augmentation alone (ablation) | CVRL · Kinetics-400 | top-1 accuracy: 33.0 -> 52.3 | Helped |
| Add both spatial and temporal augmentation instead of using spatial augmentation alone (ablation) | CVRL · Kinetics-400 | top-1 accuracy: 40.9 -> 52.3 | Helped |
| Add temporally consistent spatial augmentation (ablation) | CVRL with spatial and temporal augmentation · Kinetics-400 | top-1 accuracy: 52.3 -> 63.8 | Helped |
| Use a decreasing rather than uniform or increasing temporal-interval sampling distribution (ablation) | CVRL · Kinetics-400 | top-1 accuracy | Helped |
| Use the linear decreasing power function rather than other tested decreasing power functions (ablation) | CVRL temporal interval sampling · Kinetics-400 | top-1 accuracy | Helped |
| Replace proposed spatial augmentation with temporally consistent RandAugment (ablation) | CVRL · Kinetics-400 | top-1 accuracy: 54.2 -> 63.8 | Hurt |
| Use 3 hidden layers in the projection head instead of 0 hidden layers (ablation) | CVRL · Kinetics-400 | top-1 accuracy: 54.7 -> 63.8 | Helped |
| Use 3 hidden layers in the projection head instead of 1 hidden layer (ablation) | CVRL · Kinetics-400 | top-1 accuracy: 62.5 -> 63.8 | Helped |
| Use 3 hidden layers in the projection head instead of 2 hidden layers (ablation) | CVRL · Kinetics-400 | top-1 accuracy: 63.0 -> 63.8 | Helped |
| Increase batch size from 256 to 1024 (ablation) | CVRL · Kinetics-400 | top-1 accuracy: 60.2 -> 63.8 | Helped |
| Increase batch size from 512 to 1024 (ablation) | CVRL · Kinetics-400 | top-1 accuracy: 62.9 -> 63.8 | Helped |
| Increase batch size from 1024 to 2048 (ablation) | CVRL · Kinetics-400 | top-1 accuracy: 63.8 -> 61.8 | Hurt |
Closest papers by meaning
- Self-Supervised Spatio-Temporal Representation Learning for Videos by Predicting Motion and Appearance Statistics (CVPR 2019 · similarity 0.90)
- Contextualized Spatio-Temporal Contrastive Learning With Self-Supervision (CVPR 2022 · similarity 0.90)
- A Large-Scale Study on Unsupervised Spatiotemporal Representation Learning (CVPR 2021 · similarity 0.89)
- Spatial-Then-Temporal Self-Supervised Learning for Video Correspondence (CVPR 2023 · similarity 0.89)
- Contrastive Spatio-Temporal Pretext Learning for Self-Supervised Video Representation (AAAI 2022 · similarity 0.89)
- No More Shortcuts: Realizing the Potential of Temporal Self-Supervision (AAAI 2024 · 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.