This page provides a framework for performance evaluation of video tracking algorithms based on the evaluation protocol (PFT2) proposed in T. Nawaz, A. Cavallaro, A protocol for evaluating video trackers under real-world conditions, IEEE Transactions on Image Processing (DOI: 10.1109/TIP.2012.2228497) [video].

 

The protocol is composed of a set of trials that evaluate the robustness of trackers on a range of test scenarios representing several real-world conditions including noisy inputs, processing and communication delays, video compression and varying scene conditions such as illumination changes. The performance of the trackers is evaluated on trials using the proposed single-score evaluation measure that evaluates tracking accuracy and failure, and combines them for both summative and formative assessment. The framework uses a set of pre-defined sequences with a diversity of targets (head, vehicle, person) and challenges (occlusions, background clutter, pose changes, scale changes).

 

The software implementing the protocol and the dataset can be downloaded here.

 

We encourage you to test your tracker on PFT2 and to submit the results that will be shown here to facilitate comparisons without having to re-implement different tracking algorithms.

 

 

Contact, feedback and suggestions: pft@spevi.org