The article introduces AQATrack, a novel adaptive tracker with spatio-temporal transformers. It focuses on capturing instantaneous appearance changes using autoregressive queries and a novel attention mechanism. The proposed method aims to combine static appearance and instantaneous changes for robust tracking. Extensive experiments show significant improvements in performance across various tracking benchmarks. The article also discusses related work in visual object tracking based on spatial features and the importance of spatio-temporal information in improving discriminative ability. The method is compared with other state-of-the-art trackers, showcasing its competitive performance.
A otro idioma
del contenido fuente
arxiv.org
Ideas clave extraídas de
by Jinxia Xie,B... a las arxiv.org 03-19-2024
https://arxiv.org/pdf/2403.10574.pdfConsultas más profundas