This paper presents a novel approach to object tracking using a fully-convolutional Siamese network. The authors address the challenge of tracking arbitrary objects in videos by training a deep convolutional network to learn a similarity function that compares an exemplar image to candidate images in a search region. The network is trained end-to-end on the ILSVRC15 dataset for object detection, and the similarity function is evaluated online during tracking. The key contributions include a fully-convolutional Siamese architecture that efficiently computes the similarity function for all translated sub-windows in the search image, and the use of the large ImageNet Video dataset for training, which generalizes well to other tracking benchmarks. The proposed method achieves state-of-the-art performance in multiple benchmarks while operating at frame rates beyond real-time. The paper also discusses related work and provides experimental results to demonstrate the effectiveness of the proposed approach.This paper presents a novel approach to object tracking using a fully-convolutional Siamese network. The authors address the challenge of tracking arbitrary objects in videos by training a deep convolutional network to learn a similarity function that compares an exemplar image to candidate images in a search region. The network is trained end-to-end on the ILSVRC15 dataset for object detection, and the similarity function is evaluated online during tracking. The key contributions include a fully-convolutional Siamese architecture that efficiently computes the similarity function for all translated sub-windows in the search image, and the use of the large ImageNet Video dataset for training, which generalizes well to other tracking benchmarks. The proposed method achieves state-of-the-art performance in multiple benchmarks while operating at frame rates beyond real-time. The paper also discusses related work and provides experimental results to demonstrate the effectiveness of the proposed approach.