|Description (include details on usage, files and paper references)||PN Learning - How does TLD work?
Tracking estimates the object location as long as the object is visible. During tracking all observed patterns of the object are used to learn an object detector. The longer the object is tracked, the more accurate detector is built. The object detector searches for the object in every frame of the video. When the object re-appears with appearance seen before, the detector is able to detect it and re-initialize the tracking.
The uniqueness of TLD is the ability to combine all of these components into a tightly integrated, real-time process. The result is a general visual tracker/detector that requires only a single click for initialization and exhibits remarkable robustness.
TLD1.0 (Predator) has been released under GPL license. You can download the source code, read the installation guide and participate in a discussion group. TLD1.0 is deprecated now. See our product page for latest version of the Predator tracker.
Z. Kalal, J. Matas, and K. Mikolajczyk, “P-N Learning: Bootstrapping Binary Classifiers by Structural Constraints,” Conference on Computer Vision and Pattern Recognition, 2010.