Research map / Video / Action recognition
A Large-scale Study of Spatiotemporal Representation Learning with a New Benchmark on Action Recognition
Andong Deng, Taojiannan Yang, Chen Chen
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 |
|---|---|---|---|
| compare 2D CNN (TSM) with transformers under a matched ConvNeXt backbone | video action recognition models · BEAR 18 datasets (supervised) | macro-average accuracy: TSM 68.10 vs VideoSwin 66.48 and TimeSformer 64.27 | Improved |
| use self-supervised instead of supervised Kinetics-400 pre-training (ablation) | TSM and others · BEAR 18 datasets | macro-average accuracy: 68.10 -> 63.35 (TSM) | Hurt |
| use self-supervised instead of supervised pre-training on WLASL (ablation) | 3D Non-local network · WLASL100 | accuracy: 52.31 -> 30.56 | Hurt |
| evaluate with large viewpoint shift from Kinetics-400 (drone, egocentric, CCTV) | fine-tuned video models · UCF-Crime, UAV-Human, CharadesEgo, MECCANO | accuracy | Hurt |
| few-shot instead of full-data fine-tuning (ablation) | video models · BEAR few-shot (2,4,8,16-shot) | accuracy | Hurt |
| replace supervised pre-training by SoTA self-supervised VideoMAE in few-shot | video models · BEAR few-shot | accuracy | No clear effect |
| apply UDA methods (TA3N, CoMix) under viewpoint shift | Toyota Smarthome to MPII-Cooking · BEAR UDA | accuracy: 5.32 -> 12.63 vs supervised target 70.21 | Mixed |
| use ActionCLIP (Kinetics-400 fine-tuned) instead of CLIP for zero-shot | zero-shot evaluation · BEAR zero-shot | accuracy | Hurt |
| use NAS-based X3D versus manually designed models | video models · BEAR 18 datasets | macro-average accuracy: 61.39 vs TSM 68.10 | Hurt |
Closest papers by meaning
- STEP: Spatio-Temporal Progressive Learning for Video Action Detection (CVPR 2019 · similarity 0.90)
- AVA: A Video Dataset of Spatio-Temporally Localized Atomic Visual Actions (CVPR 2018 · similarity 0.89)
- A Closer Look at Spatiotemporal Convolutions for Action Recognition (CVPR 2018 · similarity 0.89)
- What Have We Learned From Deep Representations for Action Recognition? (CVPR 2018 · similarity 0.89)
- RubiksNet: Learnable 3D-Shift for Efficient Video Action Recognition (ECCV 2020 · similarity 0.89)
- Building a Multi-modal Spatiotemporal Expert for Zero-shot Action Recognition with CLIP (AAAI 2025 · 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.
