Detection and tracking in multicamera sports video
[Description][Related publication][Dataset][Sample results]
This page presents sample results of the developed work in the end of career project named "Detección y seguimiento en videos deportivos multicámara".
Main objective of this project is to create a system which after a
simple previous configuration is able to detect and track each player
on the court or field. Target sports to develop the system are
individual sports (e.g. tennis) and team sports (e.g. basketball and
football). Given the very different characteristics of each type of
sport, a different system for each of them has been developed.
Rafael
Martín Nieto, Detección
y seguimiento en vídeos deportivos multicámara,
Proyecto Fin de Carrera, Titulación Ingeniero de
Telecomunicación, Universidad Autónoma de Madrid, Escuela
Politécnica Superior, Septiembre 2012.
Contact Information
Rafael Martín Nieto - show email
3DLIFE ACM Multimedia dataset: Videos from cameras 1, 3 and 4 are provided here
ISSIA Soccer dataset:Videos from cameras 1 to 6 and ground-truth are provided here
Some results are presented for individual sports and team sports. For individual sports, the first video shows the tracking from each camera and the resulting trajectory. The second video shows some statistics extracted from the trajectory of the tracked player. For team sports, four videos and quantitative results are presented. The videos show the resulting fusion of facing cameras from the tracking of each camera, the region fusion from the result of facing cameras fusion, the final trajectories of each player and statistics extracted from the trajectory of the referee(to facilitate the monitoring in the videos). All the football videos use the ground-truth tracking. Precision and recall results are shown for each of the two used tracking (ground-truth tracking and base system tracking) and for each of the three types of developed fusion.