The PUT surveillance dataset is a publicly available database of color, high resolution images useful in evaluation of various algorithms in the field of video surveillance. The additional data provided with the images facilitates the evaluation of tracking, recognition and reidentification across sequences of images.

Each registered image is supplemented with an XML data file, containing information useful for evaluation. The XML format is commonly used and readily available parser libraries for a range of programming languages are available. The annotation information contains the minimum bounding boxes for all objects on the given frame (upper left and lower right corner image coordinates), as well as the object’s class (person, car, etc.).

We encourage you to check back from time to time, as the database will be constantly expanded.


Short preview of first sequence:

Short preview of second sequence:

Short preview of third sequence:

Tools used

The program for annotating the images:


This research was financed by the Polish National Science Centre grant funded according to the decision DEC-2011/03/N/ST6/03022, which is gratefully acknowledged.

More info: