Research map / Video / Action recognition
Exploring Temporally Dynamic Data Augmentation for Video 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 |
|---|---|---|---|
| Add temporally varying DynaAugment to static TrivialAugment | SlowFast-R50-16×8 · Something-Something-v2 | Top-1 accuracy: 64.7 -> 65.8 | Helped |
| Add DynaAugment to Uniformer-S-16×8 baseline training | Uniformer-S-16×8 · Kinetics-400 | Top-1 accuracy: 80.1 -> 80.9 | Helped |
| Add DynaAugment to RandAugment | S3D-G · UCF-101 | Top-1 accuracy: 82.8 -> 85.3 | Helped |
| Add DynaAugment to RandAugment | S3D-G · HMDB-51 | Top-1 accuracy: 48.5 -> 54.8 | Helped |
| Add DynaAugment to TrivialAugment | S3D-G · Diving-48 | Top-1 accuracy: 68.3 -> 72.5 | Helped |
| Add DynaAugment to UniformAugment | S3D-G · Gym288 | Top-1 per-class accuracy: 53.4 -> 56.1 | Helped |
| Use temporally dynamic augmentation instead of static augmentation (ablation) | S3D-G with RA · UCF-101 | Top-1 accuracy: 82.8 -> 85.3 | Helped |
| Replace Fourier variation with linear variation (ablation) | S3D-G with RA · UCF-101 | Top-1 accuracy: 85.3 -> 80.1 | Hurt |
| Replace Fourier variation with sinusoidal variation (ablation) | S3D-G with RA · UCF-101 | Top-1 accuracy: 85.3 -> 81.4 | Hurt |
| Replace Fourier variation with per-frame random variation (ablation) | S3D-G with RA · UCF-101 | Top-1 accuracy: 85.3 -> 76.9 | Hurt |
| Remove the random offset from Fourier Sampling (ablation) | S3D-G with RA · UCF-101 | Top-1 accuracy: 84.7 -> 85.3 | Hurt |
| Use C=1 rather than the default C=3 basis functions (ablation) | S3D-G with RA · UCF-101 | Top-1 accuracy: 84.4 -> 85.3 | Hurt |
Closest papers by meaning
- Don't Judge by the Look: Towards Motion Coherent Video Representation (ICLR 2024 · similarity 0.87)
- No More Shortcuts: Realizing the Potential of Temporal Self-Supervision (AAAI 2024 · similarity 0.86)
- Your Image is My Video: Reshaping the Receptive Field via Image-To-Video Differentiable AutoAugmentation and Fusion (CVPR 2024 · similarity 0.86)
- Minimizing Labeled, Maximizing Unlabeled: An Image-Driven Approach for Video Instance Segmentation (CVPR 2025 · similarity 0.86)
- LookWhen? Fast Video Recognition by Learning When, Where, and What to Compute (NeurIPS 2026 · similarity 0.85)
- TACO: Towards Task-Consistent Open-Vocabulary Adaptation in Video Recognition (NeurIPS 2026 · 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.