Siamese Network-Based UAV Tracking

Overview

Construct robust Siamese network-based trackers for high-performance UAV tracking. Through enhancing the anchor proposal process, feature fusion strategy, attention mechanism, etc, we have developed several robust deep learning-based trackers for UAV.

Papers with code

Related works are presented as follows:

• Proposed the anchor proposal network (APN) for adaptive anchor proposing. Alleviated the hyper-parameters in anchor-based approaches and redundent anchors in anchor-free approaches simultaneously.

  Siamese Anchor Proposal Network for High-Speed Aerial Tracking in ICRA 2021

  Onboard Real-Time Aerial Tracking with Efficient Siamese Anchor Proposal Network in IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 2021

• Integrated self-attention and cross-attention into SiamAPN, enhanced the perception ability for various scale objects of the proposed SiamAPN++. Evaluation on UAV tracking datasets and real-world onboard test demonstrate its effectiveness.

  SiamAPN++: Siamese Attentional Aggregation Network for Real-Time UAV Tracking in IROS 2021

• Introduced the hierarchical feature transformer into the Siamese framework to achieve interactive fusion of spatial and semantic cues.

  HiFT: Hierarchical Feature Transformer for Aerial Tracking in ICCV 2021

Benchmarks