IG-TRACK: IOU Guided Siamese Networks for visual object tracking (2001.09307v1)

Abstract: Recently Deep Learning based Siamese Networks with region proposals for visual object tracking becoming more popular. These networks, while testing, perform extra computations on output if trained network, to predict the bounding box. This however hampering the precision of bounding box. In this work, the authors have proposed a network guided by Intersection Over Union(IOU) while training, to predict precise bounding box. This is achieved by adding new loss function in training the network, to maximize IOU of the predicted bounding box with ground truth. While testing on VOT2018, GOT-10k tracking benchmarks,the proposed approach out-performed the base approach by more than 10% in terms of precision.