This is a demo of the crowd counting system from the CVPR 2008 paper, “Privacy Preserving Crowd Monitoring: Counting People without People Models or Tracking”. The video is available in Quicktime format (H.264).
There are more recent results from the TIP 2012 journal paper here.
Pedestrian Database (2000 frames with ground truth)
This is a demo of the crowd counting system on the pedestrian dataset (2000 frames). The system was trained on 800 frames, and tested on 1200 frames.
Crowd Counting Demo
An example video frame is shown below. The error indicator summarizes the performance of the system with respect to the ground-truth. The different colors indicate the following: (green dot) the estimate is within one standard-deviation of the groundtruth; (yellow dot) the estimate is within two standard-deviations; (red dot) the estimate is outside 2 standard-deviations. Finally, the frames marked as “training set” were used to train the GP regression model, and the region-of-interest is also highlighted. The video is sped up to twice the normal frame rate.
[mov (9.4 MB)]
Crowd Count Plot
These are plots of the crowd count over time. The gray regions are the two-standard-deviations error bars of the Gaussian process regression (i.e. the confidence of the crowd estimate). Click the image to see a larger version.
“towards” crowd
“away” crowd
Pedestrian Database (1 hour)
This is a demo of the crowd counting system on an hour-long pedestrian video. The system was trained on the 2000 frames with ground-truth, and tested on the remaining 50 minutes video. The video is sped up by 2 times, and each clip is about 3-5 minutes long.
Note that this segmentation required no reinitialization at any point, or any other type of manual supervision. The sequences contain a fair variability of traffic density, various outlying events (e.g. bicyclies, skateboarders, or small vehicles, pedestrians changing direction, etc.) and variable environmental conditions (e.g. varying clouds and shadows).
Training Set Video
 [mov (9.4 MB)] |
Test Set Video
 [mov (15 MB)] |
 [mov (16 MB)] |
 [mov (12 MB)] |
 [mov (10 MB)] |
 [mov (14 MB)] |
 |
Crowd Count Plot
This is a plot of the crowd count for all the video. The bars indicate when the system has low confidence in the count estimates (i.e. when the standard-deviation of the Gaussian process is over 3). Click the image for a larger version.
