FSRI tracks datasets that cover broadcast footage, practice sessions, and official scoring protocols. Use the catalog to plan experiments, line up pretraining corpora, and identify gaps where community collection would be valuable.
Fan-sourced practice footage with consented athlete feedback launched in 2024 for grassroots coaching analytics.
- Tasks
- Practice session tagging, Skill progression tracking, Feedback summarisation
- Modalities
- Mobile RGB video, Coach annotations, Audio notes
- Scale
- 2100 practice clips (~60 hours)
- Maintainer
- YourSkatingCoach collective
Unified annotation benchmark combining historic broadcast archives with modern call sheets for reproducible evaluation.
- Tasks
- Benchmark curation, Element verification, Technical panel simulation
- Modalities
- Broadcast RGB video, Structured annotations, Panel transcripts
- Scale
- 520 annotated programs with harmonised JSON schema
- Maintainer
- Figure Skating Research Institute
Community benchmark released for the 2024 SkatingVerse Challenge covering detection, program breakdown, and responsible score prediction.
- Tasks
- Element detection, Program diarisation, Score forecasting
- Modalities
- Broadcast RGB video, Metadata JSON
- Scale
- 320 curated programs with train/validation/test splits
- Maintainer
- SkatingVerse Consortium
Optical and inertial motion capture recordings of toe and edge jumps with synchronized force plates and expert annotations.
- Tasks
- 3D pose estimation, Biomechanical modelling, Take-off analysis
- Modalities
- Optical motion capture, IMU data, Force plate telemetry
- Scale
- 96 sessions across triples and quads
- Maintainer
- University of Calgary Human Performance Lab
500 competition videos curated for figure skating vision tasks with multilingual commentary tracks.
- Tasks
- Element detection, Speech-aligned analytics, Highlight generation
- Modalities
- Broadcast RGB video, Audio commentary, Captions
- Scale
- 500 competition programs (~80 hours)
- Maintainer
- Free Ice Skating Vision (FiSV) Initiative
271 world championship programs captured across multiple camera angles with judging metadata for AAAI 2021.
- Tasks
- Program segmentation, Performance analytics, Pose tracking
- Modalities
- Broadcast RGB video, Multi-angle video, Audio commentary
- Scale
- 271 full performances (~45 hours)
- Maintainer
- Fudan University & CASIA
1484 broadcast clips from international competitions with detailed jump phase labels and Grade of Execution references.
- Tasks
- Element classification, GOE attribution, Phase segmentation
- Modalities
- Broadcast RGB video
- Scale
- 1484 short-form clips (~12s average)
- Maintainer
- Shanghai Jiao Tong University & Tencent Youtu Lab
Early benchmark by Pirsiavash et al. (2014) with fine-grained action labels for singles programs.
- Tasks
- Action recognition, Element segmentation, Temporal detection
- Modalities
- Broadcast RGB video
- Scale
- 150 labeled programs (~18 hours)
- Maintainer
- MIT Computer Science & Artificial Intelligence Laboratory
Missing a dataset?
Help us expand the catalog with responsibly captured skating footage or structured judging data. Email board@figureskate.ai with details.