Gerard Medioni - Los Angeles CA, US Qian Yu - Plainsboro NJ, US Thang Ba Dinh - Los Angeles CA, US
Assignee:
University of Southern California - Los Angeles CA
International Classification:
G06K 9/00
US Classification:
382103
Abstract:
Technologies for object tracking can include accessing a video feed that captures an object in at least a portion of the video feed; operating a generative tracker to capture appearance variations of the object operating a discriminative tracker to discriminate the object from the object's background, where operating the discriminative tracker can include using a sliding window to process data from the video feed, and advancing the sliding window to focus the discriminative tracker on recent appearance variations of the object; training the generative tracker and the discriminative tracker based on the video feed, where the training can include updating the generative tracker based on an output of the discriminative tracker, and updating the discriminative tracker based on an output of the generative tracker; and tracking the object with information based on an output from the generative tracker and an output from the discriminative tracker.
Visual Tracking In Video Images In Unconstrained Environments By Exploiting On-The-Fly Contxt Using Supporters And Distracters
Gerard Guy Medioni - Los Angeles CA, US Thang Ba Dinh - San Jose CA, US
Assignee:
UNIVERSITY OF SOUTHERN CALIFORNIA - Los Angeles CA
International Classification:
G06K 9/00 G06T 7/20 G06K 9/46 G06K 9/62
Abstract:
The present disclosure describes systems and techniques relating to identifying and tracking objects in images, such as visual tracking in video images in unconstrained environments. According to an aspect, a system includes one or more processors, and computer-readable media configured and arranged to cause the one or more processors to: identify an object in a first image of a sequence of images, identifying one or more regions similar to the object in the first image of the sequence of images, identifying one or more features around the object in the first image of the sequence of images, preventing drift in detection of the object in a second image of the sequence of images based on the one or more regions similar to the object, and verifying the object in the second image of the sequence of images based on the one or more features.