End-to-end Deep Object Tracking with Circular Loss Function for Rotated Bounding Box
This work provides an incremental improvement in object tracking for applications requiring rotated bounding box predictions, such as autonomous driving and surveillance.
This paper addresses the problem of tracking objects with rotated bounding boxes in video streams. The authors introduce a novel end-to-end deep learning method called DOTCL, which utilizes a Transformer Multi-Head Attention architecture and a new circular loss function. DOTCL demonstrates considerable improvement in robustness over current state-of-the-art end-to-end deep learning models and outperforms state-of-the-art object tracking methods on the VOT2018 dataset in terms of expected average overlap (EAO) metric.
The task object tracking is vital in numerous applications such as autonomous driving, intelligent surveillance, robotics, etc. This task entails the assigning of a bounding box to an object in a video stream, given only the bounding box for that object on the first frame. In 2015, a new type of video object tracking (VOT) dataset was created that introduced rotated bounding boxes as an extension of axis-aligned ones. In this work, we introduce a novel end-to-end deep learning method based on the Transformer Multi-Head Attention architecture. We also present a new type of loss function, which takes into account the bounding box overlap and orientation. Our Deep Object Tracking model with Circular Loss Function (DOTCL) shows an considerable improvement in terms of robustness over current state-of-the-art end-to-end deep learning models. It also outperforms state-of-the-art object tracking methods on VOT2018 dataset in terms of expected average overlap (EAO) metric.