Dataset

The image sequences (and corresonding images) for PETS'2001 may be obtained from ftp://pets2001.cs.rdg.ac.uk/ Or alternatively access the FTP site anonymously in the usual way, i.e.:

 ftp pets2001.cs.rdg.ac.uk
 Name: ftp
 Password: <SEND YOUR E-MAIL ADDRESS AS PASSWORD>
 ftp> binary
 ftp> prompt
 ftp> cd DATASET1/TESTING/CAMERA1_JPEGS ... mget * etc

Note that it is crucial to employ "binary" transfer.

Note: there appears to be some problems accessing the ftp site using Netscape. If you are having problems, please try using Internet Explorer instead, or access via direct ftp as shown above.

Important instructions are given at the bottom of this page on processing the datasets - please read these carefully.

If you would like the dataset to be sent to you (CDROM format) please complete the form below. Please note that the main route for download is via ftp://pets.rdg.ac.uk/PETS2001/ which is hosted on a dedicated high-bandwidth RAID server. Please complete the form below only if you are unable to download via ftp

Note that due to the quantity of data involved, only datasets 1 and 2 will be sent to you by default.  However, the whole set of data is available via the ftp site.

All the datasets are multi-view (2 cameras) and are significantly more challenging than for PETS'2000 in terms of significant lighting variation, occlusion, scene activity and use of multi-view data.

The annotation (ground truth) for the datasets is available here.

Dataset 1 (training = 3064 frames, testing = 2688 frames): Moving people and vehicles.  The camera calibration may be found here.

Dataset 2 (training = 2989 frames, testing = 2823 frames): Moving people and vehicles.  The camera calibration may be found here.

Dataset 3 (training = 5563 frames, testing = 5336 frames): Moving people. This is a more challenging sequence in terms of multiple targets and significant lighting variation.  The camera calibration may be found here.

Dataset 4 (training = 6789 frames, testing = 5010 frames): Moving people and vehicles (catadioptric vision) - one narrow field of view, one panoramic.  The camera calibration may be found here.

Dataset 5 (training = 2866 frames, testing = 2867 frames): Moving vehicle (forward and rear views).  The camera calibration may be found here.

For each dataset

• the training directory contains a training set of frames in both QuickTime movie format with Motion JpegA compression and as individual Jpeg images.  The directory contains frame synchronised images for the two cameras.
• the test directory contains a training set of frames in both QuickTime movie format with Motion JpegA compression and as individual Jpeg images.   The directory contains frame synchronised images for the two cameras.

The annotation (ground truth) for the datasets is available here.
 

VERY IMPORTANT INFORMATION

The tracking results that you report in your paper submission

• should be performed using the test sequences, but the training sequences may optionally be used if the algorithms require it for learning etc.
• may be based on a single camera view of the scene (CAMERA1), or using the dual-view data (the approach used must be clearly stated in the paper).

The tracking results must be submitted along with the paper, with the tracking results generated in XML format.  This will be straightforward and should not add a significant overhead to your effort. The results you provide will be used to perform automatic performance evaluation.

The paper that you submit may be based on previously published tracking methods/algorithms (including papers submitted to the main CVPR conference).  The importance is that your paper MUST report results on tracking using the datasets supplied.

The recommendation is that:

• you attempt to track the objects (people or vehicles or both) in Datasets 1 or 2 first and report tracking results on the test sequences for these datasets in the paper.  (note that you need only use single-view data but you may use both views of the same scene (frame synchronised) if your algorithms/methods of tracking allow it.)  The pre-requisite for submitting your paper to the workshop is that you minimally report tracking results on either OR both of datasets 1 and 2.
• if you have been successfully completed the above, attempt Dataset 3 & possibly 4 (if you are interested in catadioptric vision) which are significantly longer in length and include significant lighting variation.
• Dataset 5 is primarily included for those people interested in moving cameras.  Your paper may report results on this dataset alone.