Research map / Video / Action recognition
OCSampler: Compressing Videos to One Clip With Single-Step Sampling
Jintao Lin, Haodong Duan, Kai Chen, Dahua Lin, Limin Wang
Check 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 |
|---|---|---|---|
| condense each video into one clip using a learned one-step selection policy | video recognition with ResNet · ActivityNet-v1.3 | mAP: 76.1% -> 77.2% | Improved |
| condense each video into one clip using a learned one-step selection policy | video recognition with ResNet · Mini-Kinetics | Top-1 accuracy: 72.9% -> 73.7% | Improved |
| condense videos into one clip | video recognition with ResNet · FCVID | mAP: 81.6% -> 82.7% | Improved |
| condense videos into one clip | video recognition · Mini-Sports1M | mAP: 45.0% -> 46.7% | Improved |
| use an integrated-clip reward instead of per-frame confidence (ablation) | learned frame selection on ActivityNet-v1.3 · ActivityNet-v1.3, N=4 | mAP: 74.2% -> 75.8% | Improved |
| include the uniform-sampling baseline term in the reward (ablation) | learned frame selection on ActivityNet-v1.3 · ActivityNet-v1.3, N=4 | mAP: 75.2% -> 75.8% | Improved |
| increase frame candidates from 10 to 16 (ablation) | OCSampler · ActivityNet-v1.3 | mAP: 77.2% -> 78.0% | Improved |
| increase frame candidates from 10 to 16 (ablation) | OCSampler · ActivityNet-v1.3 | GFLOPs: 25.8 -> 26.4 | Hurt |
| increase frame candidates from 10 to 24 (ablation) | OCSampler · ActivityNet-v1.3 | mAP: 77.2% -> 78.3% | Improved |
| use OCSampler-selected frames instead of baseline sampling (ablation) | ResNet and other classifiers · ActivityNet-v1.3 | mAP: 67.5 -> 75.8 | Improved |
Later papers that compared with this method
5 later accepted papers report OCSampler as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- Uni-AdaFocus: Spatial-Temporal Dynamic Computation for Video Recognition (TPAMI 2024 · better on 4 of 4 shared results)
- SpotEM: Efficient Video Search for Episodic Memory (ICML 2023 · better on 8 of 8 shared results)
- Facial Expression Recognition with Adaptive Frame Rate based on Multiple Testing Correction (ICML 2023 · better on 1 of 1 shared result)
- AdaFocusV3: On Unified Spatial-Temporal Dynamic Video Recognition (ECCV 2022 · better on 4 of 4 shared results)
- Rethinking Resolution in the Context of Efficient Video Recognition (NeurIPS 2022 · better on 3 of 3 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- AR-Net — AR-Net: Adaptive Frame Resolution for Efficient Action Recognition (ECCV 2020) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Beyond Short Clips: End-to-End Video-Level Learning With Collaborative Memories (CVPR 2021 · similarity 0.86)
- MGSampler: An Explainable Sampling Strategy for Video Action Recognition (ICCV 2021 · similarity 0.86)
- PackForcing: Short Video Training Suffices for Long Video Sampling and Long Context Inference (ECCV 2026 · similarity 0.86)
- Frozen CLIP Models Are Efficient Video Learners (ECCV 2022 · similarity 0.85)
- One Token per Highly Selective Frame: Towards Extreme Compression for Long Video Understanding (NeurIPS 2025 · similarity 0.84)
- ODTrack: Online Dense Temporal Token Learning for Visual Tracking (AAAI 2024 · similarity 0.84)
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.
