Study on Deep Learning and Its Application in Visual Tracking

摘要

Inspired by recent advances in deep learning, this paper reviews the deep learning methodologies and its applications in object tracking. To overcome the complexity and low-efficiency of existing full-connected deep learning based tracker, we use a novel convolutional deep belief network (CDBN) with convolution, weights sharing and pooling to have much fewer parameters, in addition to gain translation invariance which would benefit the tracker performance. Empirical evaluation demonstrates our CDBN based tracker outperforms several state-of-the-art methods on an open tracker benchmark.