TRICTRAC Video Dataset: Public HDTV Synthetic Soccer Video Sequences With Ground Truth. X. Desurmont 1 , J-B. Hayet 2 , J-F. Delaigle 1 , J. Piater 2 , B. Macq 3 1 Multitel ASBL, Parc Initialis – Avenue Copernic, 1, B-7000, Mons, Belgium 2 Institut Montefiore, University of Liège, Building B28, B-4000 Liège, Belgium 3 Communications and Remote Sensing Laboratory, UCL, Louvain-la-neuve,Belgium Abstract. Object tracking in video sequences is an important task in many applications such as video surveillance, traffic monitoring, marketing and sport analysis. In order to enhance these technologies, an objective performance evaluation is needed. This evaluation requires to test the system with a given dataset and compare the output with the ground truth. One of the contributions of the TRICTRAC project is the supply to the video processing community of synthetic, high-definition video content of Pan-Tilt-Zoom (PTZ) cameras with 3D ground truth including the parameters of the cameras and the mobile objects. This paper presents this novel dataset. 1. INTRODUCTION Testing video-analysis algorithms is very important in the academic and industrial communities. To enable performance evaluation, multiple steps are required. First, video sequences must be available. Secondly, the Video Content Analysis (VCA) technique to be evaluated must generate results on the test sequences. Thirdly, Ground Truth (GT) needs to be available. Then, the GT needs to be compared with the generated results, requiring an unambiguous definition of the metrics. Finally, the evaluation results are combined for each video sequence to be presented to the user. The Performance Evaluations of Tracking and Surveillance (PETS) workshop deals with these issues since 2000. It proposes datasets for surveillance, sports, smart meetings, etc. However, there are no datasets for Pan-Tilt-Zoom (PTZ) cameras with full camera ground truth available. In this paper, we focus on the synthetic video datasets and their associated GT produced during the TRICTRAC [1] project. This is a scientific project that aims to build a tracking system for networks of multiple PTZ cameras. The three principle steps are to track objects in each video stream, to recover PTZ parameters in real time, and to merge the recovered camera and object parameters to build a 3D representation of the dynamic scene. The paper is organized as follows: Section 2 discusses related work. Section 3 presents the modelling of the scene, objects, cameras and scenarios, and Section 4 explains the synthetic video and the ground truth format.