About This Topic
The detection of objects of interest in an image or video is a challenging task. Recently, methods for the automatic learning of object models have been introduced that enable the detection of even complex objects, such as animals.
Elephants are complex objects to detect automatically. In this thesis/practical we investigate a novel parameter-less method that automatically extracts image segments from a given set of training images that correspond to elephants in a wildlife video. Based on the segments, a visual dictionary is built that is later used for elephant detection.