Online Object Tracking based on Generative Adversarial Networks

Object Tracking is one of the essential steps in computer vision applications such as human-computer interaction, video surveillance, and augmented reality. Visual object tracking remains one of the most challenging tasks in object tracking due to many variations of conditions. In this paper, we utilize Residual Network ResNet as a feature extractor and address the limitations of object tracking using Generative Adversarial Networks. We describe a generative adversarial network to improve object tracker’s classifier during the online training phase. The network generates adaptive features to augment the positive samples detected by the convolutional layer of the tracker’s model in order to improve the model’s classifier. The network is integrated with Generative Adversarial Networks tracker and the object tracking results are presented.