Research map / Video / Video segmentation and tracking
VideoMoCo: Contrastive Video Representation Learning With Temporally Adversarial Examples
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 adversarial frame dropout with k=8 (ablation) | MoCo with R(2+1)D · UCF101 | top-1 accuracy: 75.9 -> 77.8 | Helped |
| Add adversarial frame dropout with k=8 (ablation) | MoCo with R(2+1)D · HMDB51 | top-1 accuracy: 46.2 -> 49.1 | Helped |
| Replace adversarial frame selection with random dropout (ablation) | MoCo with R(2+1)D · UCF101 | top-1 accuracy: 75.9 -> 74.1 | Hurt |
| Replace adversarial frame selection with random dropout (ablation) | MoCo with R(2+1)D · HMDB51 | top-1 accuracy: 46.2 -> 43.6 | Hurt |
| Vary the number of adversarially dropped frames from k=8 to k=4, 12, 16, or 24 (ablation) | VideoMoCo with R(2+1)D · UCF101 | top-1 accuracy: 77.8; 76.4; 77.1; 75.2; 73.4 | Mixed |
| Vary the number of adversarially dropped frames from k=8 to k=4, 12, 16, or 24 (ablation) | VideoMoCo with R(2+1)D · HMDB51 | top-1 accuracy: 49.1; 48.6; 45.6; 40.3; 39.0 | Mixed |
| Add temporal decay (t=0.99999) (ablation) | VideoMoCo with adversarial dropout k=8 · UCF101 | top-1 accuracy: 77.8 -> 78.7 | Helped |
| Remove adversarial dropout while retaining temporal decay (ablation) | VideoMoCo with R(2+1)D · UCF101 | top-1 accuracy: 78.7 -> 76.1 | Hurt |
| Set temporal decay to t=0.999 (ablation) | VideoMoCo with adversarial dropout · UCF101 | top-1 accuracy: 77.8 -> 73.4 | Hurt |
| Set temporal decay to t=0.9999 (ablation) | VideoMoCo with adversarial dropout · UCF101 | top-1 accuracy: 77.8 -> 75.2 | Hurt |
| Set temporal decay to t=0.999999 (ablation) | VideoMoCo with adversarial dropout · UCF101 | top-1 accuracy: 77.8 -> 78.3 | Helped |
| Set temporal decay to t=0.9999999 (ablation) | VideoMoCo with adversarial dropout · UCF101 | top-1 accuracy: 77.8 -> 77.9 | Helped |
Closest papers by meaning
- Momentum Contrast for Unsupervised Visual Representation Learning (CVPR 2020 · similarity 0.89)
- Spatiotemporal Contrastive Video Representation Learning (CVPR 2021 · similarity 0.87)
- Slow and Steady Feature Analysis: Higher Order Temporal Coherence in Video (CVPR 2016 · similarity 0.87)
- Multi-Scale Contrastive Learning for Video Temporal Grounding (AAAI 2025 · similarity 0.85)
- DejaVid: Encoder-Agnostic Learned Temporal Matching for Video Classification (CVPR 2025 · similarity 0.85)
- Temporally Consistent Object-Centric Learning by Contrasting Slots (CVPR 2025 · similarity 0.85)
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.