Research map / Video / Action recognition
FineGym: A Hierarchical Video Dataset for Fine-Grained Action Understanding
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 |
|---|---|---|---|
| increase the number of sampled training frames (ablation) | TSN · Gym99 | Top-1: 35.46 -> 78.82 | Helped |
| use optical flow rather than RGB appearance | TSN · Gym99 | Mean: 61.4 -> 75.6 | Helped |
| use optical flow rather than RGB appearance | TSN · VT, 6cls | Mean: 27.8 -> 23.1 | Hurt |
| replace TSN's temporal averaging with longer-term temporal reasoning | TRN · Gym288 | Mean: 26.5 -> 33.1 | Helped |
| use temporal shifting instead of sparse segment consensus | TSM · Gym288 | Mean: 26.5 -> 34.8 | Helped |
| shuffle input frames at test time (ablation) | TRN on UB-circles · UB-circles | performance | Hurt |
| test TSM with a frame count very different from training (ablation) | TSM · FineGym | performance | Hurt |
| pre-train on coarse-grained action datasets | I3D and Non-local networks · FineGym element recognition | recognition accuracy | Mixed |
| localize sub-actions rather than complete actions | SSN · GymFine | Average mAP@tIoU: 49.4 -> 9.6 | Hurt |
Closest papers by meaning
- ActionAtlas: A VideoQA Benchmark for Domain-specialized Action Recognition (NeurIPS 2024 · similarity 0.90)
- FineSports: A Multi-person Hierarchical Sports Video Dataset for Fine-grained Action Understanding (CVPR 2024 · similarity 0.89)
- SeFAR: Semi-supervised Fine-grained Action Recognition with Temporal Perturbation and Learning Stabilization (AAAI 2025 · similarity 0.89)
- FineParser: A Fine-grained Spatio-temporal Action Parser for Human-centric Action Quality Assessment (CVPR 2024 · similarity 0.88)
- Storyboard-guided Alignment for Fine-grained Video Action Recognition (NeurIPS 2025 · similarity 0.87)
- Unsupervised Action Segmentation via Fast Learning of Semantically Consistent Actoms (AAAI 2024 · similarity 0.87)
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.