Fis-V: Figure Skating Video Scoring
Automatic figure skating scoring as video regression. Annotators mark the technical elements and record the Total Element Score (TES) and Total Program Component Score (PCS) for a ladies' short program (Xu et al., IEEE TCSVT 2020).
About this dataset
Fis-V is a dataset for automatically scoring figure skating performances, introduced by Chengming Xu, Yanwei Fu, and colleagues (IEEE Transactions on Circuits and Systems for Video Technology, 2020). It frames scoring as video regression: predict the referee scores a skater would receive from the full performance video.
The videos are ladies' singles short-program performances from international competitions. Each video is annotated with two official referee scores: the Total Element Score (TES), reflecting the difficulty and execution of technical elements, and the Total Program Component Score (PCS), reflecting skating skills, transitions, performance, composition, and music interpretation. Scores are the official values from a panel of nine referees.
The dataset contains 500 videos from 149 professional skaters, each roughly two minutes fifty seconds long. Every video carries the two score labels (TES and PCS); the paper also evaluates on the earlier MIT-Skate dataset.
The Potato config below reproduces the scoring task with a video segment scheme for the technical elements plus fields for the Total Element Score and Total Program Component Score. Use it to collect human scores for figure skating or as a template for scored-sport video assessment.
- Released
- IEEE TCSVT 2020
- Sport
- Figure skating (ladies' short program)
- Videos
- 500
- Skaters
- 149
- Score labels
- TES + PCS (per video)
- Referees
- 9
Configuration Fileconfig.yaml
This Potato config reproduces the annotation task. Save it as config.yaml and run potato start config.yaml to try it.
# Fis-V - Figure Skating Video Scoring
# Based on Xu, Fu, Zhang, Chen, Jiang & Xue, IEEE TCSVT 2020
# Paper: https://arxiv.org/abs/1802.02774
# Dataset: https://github.com/loadder/MS_LSTM
#
# Fis-V frames figure skating scoring as video regression: predict the
# referee scores a skater would receive from the full performance video.
# Each ladies' singles short-program video carries two official scores:
# the Total Element Score (TES) and the Total Program Component Score (PCS).
#
# This Potato config reproduces the scoring: watch the performance, mark the
# technical elements, and record the TES and PCS.
annotation_task_name: "Fis-V - Figure Skating Video Scoring"
task_dir: "."
data_files:
- sample-data.json
item_properties:
id_key: "id"
text_key: "video_url"
output_annotation_dir: "annotation_output/"
output_annotation_format: "json"
annotation_schemes:
# Step 1: mark the technical elements on the timeline
- name: "elements"
description: "Mark each technical element (jump, spin, step sequence) on the timeline."
annotation_type: "video_annotation"
mode: "segment"
labels:
- name: "jump"
color: "#6366F1"
key_value: "j"
- name: "spin"
color: "#14B8A6"
key_value: "s"
- name: "step_sequence"
color: "#F59E0B"
key_value: "t"
frame_stepping: true
show_timecode: true
playback_rate_control: true
video_fps: 25
# Step 2: Total Element Score (TES)
- annotation_type: text
name: total_element_score
description: "Total Element Score (TES) — difficulty and execution of technical elements."
# Step 3: Total Program Component Score (PCS)
- annotation_type: text
name: total_program_component_score
description: "Total Program Component Score (PCS) — skating skills, transitions, performance, composition, interpretation."
# Step 4: notes on the elements performed
- annotation_type: text
name: element_notes
description: "Optional notes on the elements performed and their quality."
annotation_instructions: |
You will see a ladies' singles short-program figure skating performance.
1. Mark each technical element on the timeline (jump, spin, step sequence).
2. Record the Total Element Score (TES), which reflects the difficulty and
execution of the technical elements.
3. Record the Total Program Component Score (PCS), which reflects skating
skills, transitions, performance, composition, and music interpretation.
4. Optionally add notes on the elements performed.
In the official competition these two scores come from a panel of nine
referees; here you record the values that summarize the performance.
html_layout: |
<div style="padding: 15px; max-width: 800px; margin: auto;">
<div style="background: #eef2ff; border: 1px solid #c7d2fe; border-radius: 8px; padding: 10px 14px; margin-bottom: 12px; font-size: 13px; color: #3730a3;">
Skating performance {{clip_id}} — mark the elements and record TES and PCS.
</div>
<video controls style="width: 100%; border-radius: 8px;" src="{{video_url}}"></video>
</div>
allow_all_users: true
instances_per_annotator: 40
annotation_per_instance: 2
allow_skip: true
skip_reason_required: false
Sample Datasample-data.json
[
{
"id": "fisv_001",
"clip_id": "WC-SP-014",
"video_url": "https://example.com/fisv/clip_001.mp4",
"text": "Ladies' short program: opening triple-triple combination jump."
},
{
"id": "fisv_002",
"clip_id": "GPF-SP-032",
"video_url": "https://example.com/fisv/clip_002.mp4",
"text": "Ladies' short program: layback spin into a step sequence."
}
]
// ... and 3 more itemsGet This Design
Clone or download from the repository
Quick start:
git clone https://github.com/davidjurgens/potato-showcase.git cd potato-showcase/video/action-recognition/fisv-figure-skating-scoring potato start config.yaml
Dataset & paper
Xu et al., IEEE TCSVT 2020
Citation (BibTeX)
@article{xu2020learning,
title={Learning to Score Figure Skating Sport Videos},
author={Xu, Chengming and Fu, Yanwei and Zhang, Bing and Chen, Zitian and Jiang, Yu-Gang and Xue, Xiangyang},
journal={IEEE Transactions on Circuits and Systems for Video Technology},
volume={30},
number={12},
pages={4578--4590},
year={2020},
doi={10.1109/TCSVT.2019.2927118}
}Details
Annotation Types
Domain
Use Cases
Tags
Found an issue or want to improve this design?
Open an IssueRelated Designs
FineDiving: Procedure-aware Action Quality Assessment
Fine-grained diving action quality assessment. Annotators mark a dive and its consecutive sub-action steps on the timeline, then record the dive type, difficulty degree, and scores (Xu et al., CVPR 2022).
ActivityNet Captions: Dense Video Captioning Dataset
ActivityNet Captions pairs 20k untrimmed videos with 100k temporally localized sentence descriptions for dense-captioning research. This Potato config reproduces the segment-and-describe workflow.
VSTAR Video-grounded Dialogue
Video-grounded dialogue annotation. Annotators watch videos and answer questions requiring situated understanding, write dialogue turns grounded in specific video moments, and mark relevant temporal segments.