Women’s Soccer Penalty Kick Data Captured by Computer Vision Models

Published: 18 February 2026| Version 2 | DOI: 10.17632/brx9bsxnpx.2
Contributors:
,

Description

The penalty_pose_keypoints.csv dataset contains frame-level 2D pose-tracking data for women’s soccer penalty kicks, extracted from standard video footage using a reproducible YOLOv8 pose pipeline. Each observation corresponds to a single video frame (mean FPS = 25), spanning the run-up from the start of a penalty kick to the final frame at ball contact. The dataset includes both identifier and outcome labels for each kick (e.g., kick id, kick outcome, kick direction, player footedness, and camera direction), as well as tracking metadata for the detected kicker. For each frame, the file records the smoothed bounding-box coordinates and 12 COCO body joints (shoulders through ankles; indices 5–16) as 2D pixel coordinates. Joint locations are provided in two coordinate systems: (1) absolute pixel positions in the image (origin at the top-left), and (2) box-centered coordinates created by subtracting the axes' center points to support cross-view comparison and normalization. The published dataset contains 11,140 frames from 133 penalties recorded during collegiate women's soccer training sessions and games.

Files

Steps to reproduce

To reproduce this dataset, future users must (a) access the YOLO-v8 package in python, (b) run the model on the clips contained in the Women_Players_Videos.zip file in the Instructions and Code folder, and (c) save the outputted data in a usable file format. A README is provided in the Instructions and Code folder.

Institutions

Categories

Computer Vision, Sport, Soccer, Data Analytics

Licence