Adaptive and Background-Aware Vision Transformer for Real-Time UAV Tracking

Shuiwang Li1; Yangxiang Yang1; Dan Zeng2; Xucheng Wang1

10.1109/ICCV51070.2023.01286

2023-10-04

International Conference on Computer Vision

1550-5499

979-8-3503-0719-1

2023 IEEE/CVF International Conference on Computer Vision (ICCV)

13943-13954

October 2 - 6, 2023

Paris, France

10662 LOS VAQUEROS CIRCLE, PO BOX 3014, LOS ALAMITOS, CA 90720-1264 USA

IEEE COMPUTER SOC

While discriminative correlation filters (DCF)-based trackers prevail in UAV tracking for their favorable efficiency, lightweight convolutional neural network (CNN)-based trackers using filter pruning have also demonstrated remarkable efficiency and precision. However, the use of pure vision transformer models (ViTs) for UAV tracking remains unexplored, which is a surprising finding given that ViTs have been shown to produce better performance and greater efficiency than CNNs in image classification. In this paper, we propose an efficient ViT-based tracking framework, Aba-ViTrack, for UAV tracking. In our framework, feature learning and template-search coupling are integrated into an efficient one-stream ViT to avoid an extra heavy relation modeling module. The proposed Aba-ViT exploits an adaptive and background-aware token computation method to reduce inference time. This approach adaptively discards tokens based on learned halting probabilities, which a priori are higher for background tokens than target ones. Extensive experiments on six UAV tracking benchmarks demonstrate that the proposed Aba-ViTrack achieves state-of-the-art performance in UAV tracking. Code is available at https://github.com/xyyang317/Aba-ViTrack.

Couplings Training Representation learning Target tracking Object detection Benchmark testing Autonomous aerial vehicles

通讯

英语

CPCI-S

Guangxi Science and Technology Base and Talent Special Project[GKAD22035127] ; National Natural Science Foundation of China["62206123","62066042","62176170","61971005"] ; Sichuan Province Key Research and Development Project[2020YJ0282]

Computer Science ; Imaging Science & Photographic Technology

Computer Science, Artificial Intelligence ; Computer Science, Theory & Methods ; Imaging Science & Photographic Technology

WOS:001169499006038

人工提交

1.College of Information Science and Engineering , Guilin University of Technology, China
2.Research Institue of Trustworthy Autonomous Systems, Southern University of Science and Technology, China

Shuiwang Li,Yangxiang Yang,Dan Zeng,et al. Adaptive and Background-Aware Vision Transformer for Real-Time UAV Tracking[C]. 10662 LOS VAQUEROS CIRCLE, PO BOX 3014, LOS ALAMITOS, CA 90720-1264 USA:IEEE COMPUTER SOC,2023:13943-13954.