=== TRAFFIC CLUSTERING DATABASE ===

Statistical Visual Computing Lab
University of California, San Diego

Antoni Chan  [abchan@ucsd.edu]
http://www.svcl.ucsd.edu/


-- NOTES --------------------------
This is the database of highway traffic used for video clustering in [1].  The 
video was taken over two days from a stationary camera overlooking I-5 in 
Seattle, WA.  The pre-processed video used in [1] (i.e. cropped, downsized, and
registered) are provided in a MATLAB file, "traffic_patches_reg.mat".  In
addition, the original videos are also included in the "video" directory.  Note
that the first frame of the original video is corrupted with another video
signal, so when processing the original video you will have to start from the 
2nd frame.  The video is provided courtesy of Washington State Department of 
Transportation [http://www.wsdot.wa.gov/].  Please reference [1] if you use this
database.

More information about each video is provided in "info.txt".  The videos were 
labeled manually as light, medium, and heavy traffic, which correspond 
respectively to free-flowing traffic, traffic at reduced speed, and stopped or 
very slow speed traffic.  This database is a based on the traffic classification
database used in [2,3].

Data from the speed-sensor located on the highway is also included in the MATLAB
file "sensordata.mat", which has the following variables:
     sensor_speed     - speed measurements
     sensor_timestamp - timestamp of each measurement
     video_class      - ground-truth class for each video
     video_timestamp  - timestamp for each video in database
NOTE: Unfortunately, a precise comparison between the speed measurements and the
video is not possible because the data originate from two separate systems, and
the video data is not time-stamped with fine enough precision. Nonetheless, it
is still possible to examine the correspondences between the speed data and the
video clustering (e.g see [1]).


-- FILES ---------------------------
  README.txt              - this file
  ImageMaster.mat         - list of videos in database
  traffic_patches_reg.mat - traffic patches, registered by hand
  video/                  - original traffic videos
  info.txt                - more information about each video
  sensordata.mat          - speed sensor data


-- LOCATION ------------------------
Source:    http://www.wsdot.wa.gov/
City:      Seattle, WA.
Location:  I-5 S 188th St.
Direction: looking south
Traffic:   Southbound traffic
Date:      08/05/2004 to 08/06/2004
Loop:      152.2


-- REFERENCES ----------------------
[1] A. B. Chan and N. Vasconcelos, "Modeling, clustering, and segmenting video 
    with mixtures of dynamic textures"  IEEE Trans. on Pattern Analysis and 
    Machine Intelligence, vol. 30(5), pp. 909-926, May 2008. 

[2] A. B. Chan and N. Vasconcelos, "Probabilistic Kernels for the Classification
    of Auto-Regressive Visual Processes".  Proceedings of IEEE Conference on 
    Computer Vision and Pattern Recognition, San Diego, 2005. 

[3] A. B. Chan and N. Vasconcelos, "Classification and Retrieval of Traffic 
    Video using Auto-Regressive Stochastic Processes".  Proceedings of 2005
    IEEE Intelligent Vehicles Symposium, Las Vegas, June 2005.


VERSIONS:
04-02-2008 - initial version

================================================================================
