Jump-Aware: Player Position Rectification and Identification in Dynamic Sports Using Jump Event Spotting

SportsJumpMotion Dataset for the CVPR CVSports 2025 Workshop – contains detailed information about the dataset files and structure.

Overview

The SportsJumpMotion dataset is designed for player position rectification and identification in dynamic sports. It includes comprehensive data to support various computer vision tasks in sports analytics, such as:

• Player Detection and Tracking:
  Tracklets of cropped bounding boxes extracted from original video frames.
• Event Analysis:
  Detailed annotations of jump events and jersey number information at the tracklet level.
• Binary Segmentation:
  Binary segmentation masks provided as .png files at original frame resolution.
• Spatial Analysis:
  Original video frames and optional homography .npy files for advanced geometric transformations.
• Jersey Number Legibility:
  JSON files that indicate jersey number visibility on a frame-by-frame basis.

Directory Details

• event_annotations/
  Contains annotation files for jump events and jersey number annotations at the tracklet level. These files provide the temporal and spatial details of each jump event, along with corresponding jersey numbers.

• images/
  Contains tracklets of cropped bounding boxes for players. Each sequence in this folder represents a series of images extracted from the original video frames, focusing on individual players.

• images_binary/
  Contains binary segmentation .png files that match the original frame resolution. These masks are useful for extracting and analyzing player regions or bounding boxes in the images.

• json_legibility/
  Contains JSON files that specify jersey number visibility for each frame within a tracklet. A value of 1 indicates that the jersey number is visible, while 0 indicates it is not.

• video/
  Contains the original video frames or video clips from which the tracklets are derived. This folder may also include optional homography .npy files that can be used for mapping or transforming coordinates in spatial analysis tasks.

Accessing the Dataset

The full dataset is hosted on OneDrive.

Steps to Access:

1. Please send an email to [email protected] to request access.
2. Download the entire dataset folder once access is granted.
3. Extract the files locally.
4. Follow the usage instructions provided below.

Usage

• Event Annotation Analysis:
  Use the files in the event_annotations/ folder to identify jump events and view detailed jersey number annotations at the tracklet level.

• Player Tracklets:
  The images/ folder contains sequences of cropped bounding boxes for players.

• Binary Segmentation:
  The images_binary/ folder provides binary segmentation masks that can assist in refining player localization and segmentation analyses.

• Jersey Number Legibility:
  The JSON files in the json_legibility/ folder indicate jersey number visibility (1 for visible, 0 for invisible) on a per-frame basis, which is valuable for evaluating jersey detection performance.

• Video Reference and Homography:
  The video/ folder offers original video frames/clips for reference. Additionally, optional homography .npy files support spatial transformations and advanced analysis.

Citation

If you use this dataset in your research, please cite it as follows:

@InProceedings{Oo_2025_CVPR,
    author    = {Oo, Yin May and Jamsrandorj, Ankhzaya and Chao, Vanyi and Nguyen, Hoang Quoc and Hwang, Yewon and Mun, Kyung-Ryoul and Kim, Jinwook},
    title     = {Jump-Aware: Player Position Rectification and Identification in Dynamic Sports Using Jump Event Spotting},
    booktitle = {Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR) Workshops},
    month     = {June},
    year      = {2025},
    pages     = {5935-5944}
}

Contact

For further information or questions, please contact:

Yin May Oo
Email: [[email protected]]