PETS 2000 sequence

The input video and several processing stages can be viewed online below. The top two windows show the input and output sequence respectively. The large frame with four subdivisions shows various processing stages simultaneously. From top left to bottom right, the subframes show the raw input sequence, the detected foreground pixels highlighted in red hue, the most likely background estimate (this changes only slowly when the scene geometry changes), and the processed output video. Tracked objects are labelled with numbers and surrounded by bounding boxes. Green text and bounding boxes signify that the tracking Kalman filter received new information during this frame. A black label signifies that no new data was received and that the Kalman filter hypothesised the current object position solely from past information.
Input Output

Get the Flash Player to see this player.

Get the Flash Player to see this player.

Multiple stages simultaneously

Get the Flash Player to see this player.

If you are having difficulties viewing these sequences, you can download the videos directly: input output multiple stages.
We would like to thank the University of Reading for making the image sequence available. The original sequence can be found at their website.