Research map / Video / Action recognition
Omni-sourced Webly-supervised Learning for Video Recognition
Haodong Duan, Yue Zhao, Yuanjun Xiong, Wentao Liu, Dahua Lin
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 OmniSource web data | TSN-3seg ResNet50 ImageNet · Kinetics-400 val | Top-1 accuracy: 70.6 -> 73.6 | Improved |
| add OmniSource web data | SlowOnly-4x16 R50 from scratch · Kinetics-400 val | Top-1 accuracy: 72.9 -> 76.8 | Improved |
| add OmniSource web data | SlowOnly-4x16 R50 ImageNet · Kinetics-400 val | Top-1 accuracy: 73.8 -> 76.6 | Improved |
| add OmniSource web data | irCSN-152 IG-65M · Kinetics-400 val | Top-1 accuracy: 82.6 -> 83.6 | Improved |
| add OmniSource web data | SlowOnly-8x8 R101 from scratch · Kinetics-400 val | Top-1 accuracy: 76.3 -> 80.4 | Improved |
| skip teacher filtering of web data (ablation) | TSN joint training · Kinetics-400 val | Top-1 accuracy: 70.6 -> 67.4 | Hurt |
| add Google and Instagram images with teacher filtering (ablation) | TSN-3seg R50 · Kinetics-400 val | Top-1 accuracy: 70.6 -> 72.0 | Improved |
| add untrimmed videos (ablation) | TSN-3seg R50 · Kinetics-400 val | Top-1 accuracy: 70.6 -> 71.7 | Improved |
| use all sources combined (ablation) | TSN-3seg R50 · Kinetics-400 val | Top-1 accuracy: 70.6 -> 73.6 | Improved |
| use class-agnostic perspective warp instead of still replication (ablation) | SlowOnly-4x16 R50 · Kinetics-400 val | Top-1 accuracy: 74.1 -> 74.5 | Improved |
| use random translation to turn images into clips (ablation) | SlowOnly-4x16 R50 · Kinetics-400 val | Top-1 accuracy: 73.8 -> 73.7 | Hurt |
| add mixup (ablation) | SlowOnly-4x16 R50 from scratch · Kinetics-400 val | Top-1 accuracy: 72.9 -> 73.3 | Improved |
Closest papers by meaning
- Data Distillation: Towards Omni-Supervised Learning (CVPR 2018 · similarity 0.88)
- Webly Supervised Fine-Grained Recognition: Benchmark Datasets and an Approach (ICCV 2021 · similarity 0.86)
- OmniViD: A Generative Framework for Universal Video Understanding (CVPR 2024 · similarity 0.84)
- OmniSegmentor: A Flexible Multi-Modal Learning Framework for Semantic Segmentation (NeurIPS 2025 · similarity 0.83)
- Stable Mean Teacher for Semi-supervised Video Action Detection (AAAI 2025 · similarity 0.83)
- Is ImageNet worth 1 video? Learning strong image encoders from 1 long unlabelled video (ICLR 2024 · similarity 0.83)
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.
