A Video Tracking Algorithm Based on Neural Network

Yang Luyu

Abstract: Target tracking is an important research direction in computer vision. In order to realize target tracking, this paper proposes a target tracking algorithm based on deep neural network. The basic framework of the algorithm consists of a Siamese network. The main innovation is that the attention mechanism module is added to the Siamese network model. The attention mechanism module can make the extracted features more refined, so that the tracking effect is more accurate. The model proposed in this paper achieves excellent tracking results on both OTB2015 and UAV123 test datasets.

Keywords: Target tracking, Deep neural network, Siamese network, Attention mechanism