Robust Tracking via Convolutional Networks without Learning

Deep networks have been successfully applied to visual tracking by learning a generic representation offline from numerous training images. However the offline training is time-consuming and the learned generic representation may be less discriminative for tracking specific objects. In this paper we present that, even without learning, simple convolutional networks can be powerful enough to develop a robust representation for visual tracking. In the first frame, we randomly extract a set of normalized patches from the target region as filters, which define a set of feature maps in the subsequent frames. These maps measure similarities between each filter and the useful local intensity patterns across the target, thereby encoding its local structural information. Furthermore, all the maps form together a global representation, which maintains the relative geometric positions of the local intensity patterns, and hence the inner geometric layout of the target is also well preserved. A simple and effective online strategy is adopted to update the representation, allowing it to robustly adapt to target appearance variations. Our convolution networks have surprisingly lightweight structure, yet perform favorably against several state-of-the-art methods on a large benchmark dataset with 50 challenging videos.