Ball Detection Datasets

Ball detection datasets comprises three diffrent publicly available datasets for soccer, basketball, and volleyball sports. These datasets consist of approximately 34000 frames. They are labeled manually, frame by frame, for the purpose of academic studies in ball detection by the members of Image Processing Laboratory (IPL) of Sharif University of Technology. Ball coordinates in each frame are available in csv format. For more information please contact skasaei[at]sharif[dot]edu.




For citation purposes, you can use our publication.

Ball Locations

Dataset Sport Link Resolution No. of Fixed Cameras Total No. of Frames Ball Locations (.csv)
ISSIA CNR Soccer 1920x1080 6 3000 Cam 1
3000 Cam 2
3000 Cam 3
3000 Cam 4
3000 Cam 5
3000 Cam 6
APIDIS Basketball 1600x1200 5 1235 Cam 1
1222 Cam 2
1279 Cam 4
1288 Cam 6
1323 Cam 7
YACVID Volleyball 1920x1080 1 10000 graz-arbesbach_5


  • R. Fahmi, M. Zandigohar, and S. Kasaei. Ball Tracking Framework based on Deep Learners and Eigenballs, (to appear), 2016.