Research map / Video / Action recognition
"Spotting Temporally Precise, Fine-Grained Events in Video"
James Hong, Haotian Zhang, Michaël Gharbi, Matthew Fisher, Kayvon Fatahalian
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 GRU on top of per-frame features (ablation) | E2E-Spot · FineGym-Full | mAP (delta=1): -14.5 (32.9 vs default) | Improved |
| add GRU on top of per-frame features (ablation) | E2E-Spot · FS-Perf | mAP (delta=1): -7.2 (77.9 vs default) | Improved |
| add GSM temporal shift in the 2D CNN (ablation) | E2E-Spot · FS-Comp | mAP (delta=1): -5.5 (75.5 vs 81.0 without GSM) | Improved |
| add GSM temporal shift in the 2D CNN (ablation) | E2E-Spot · FS-Perf | mAP (delta=1): +0.4 (85.6 without GSM) | No clear effect |
| include both GSM and GRU instead of neither (ablation) | E2E-Spot · FS-Comp | mAP (delta=1): 26.9 -> 81.0 | Improved |
| use GSM instead of TSM fixed shift (ablation) | E2E-Spot · FS-Comp | mAP (delta=1): 78.6 -> 81.0 | Improved |
| use longer clips of 100 frames instead of 8 (ablation) | E2E-Spot · FS-Comp | mAP (delta=1): 73.7 -> 81.0 | Improved |
| use longer clips of 100 frames instead of 8 (ablation) | E2E-Spot · Tennis | mAP (delta=1): 95.8 -> 96.1 | No clear effect |
| use clip length 250 instead of 100 (ablation) | E2E-Spot · FineGym-Full | mAP (delta=1): +1.1 | No clear effect |
| use 224 px input instead of 112 px (ablation) | E2E-Spot · Tennis | mAP (delta=1): 88.5 -> 96.1 | Improved |
| replace GRU with MS-TCN (ablation) | E2E-Spot · FS-Comp | mAP (delta=1): 81.0 -> 77.6 | Hurt |
| replace GRU with ASFormer (ablation) | E2E-Spot · FS-Comp | mAP (delta=1): 81.0 -> 68.4 | Hurt |
Closest papers by meaning
- Precise Event Spotting in Sports Videos: Solving Long-Range Dependency and Class Imbalance (CVPR 2025 · similarity 0.89)
- A Context-Aware Loss Function for Action Spotting in Soccer Videos (CVPR 2020 · similarity 0.89)
- FineSports: A Multi-person Hierarchical Sports Video Dataset for Fine-grained Action Understanding (CVPR 2024 · similarity 0.87)
- MomentSeeker: A Task-Oriented Benchmark For Long-Video Moment Retrieval (NeurIPS 2025 · similarity 0.86)
- $F^3Set$: Towards Analyzing Fast, Frequent, and Fine-grained Events from Videos (ICLR 2025 · similarity 0.85)
- FineParser: A Fine-grained Spatio-temporal Action Parser for Human-centric Action Quality Assessment (CVPR 2024 · 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.
