English

BiFormer: Vision Transformer with Bi-Level Routing Attention

Computer Vision and Pattern Recognition 2023-03-16 v1

Abstract

As the core building block of vision transformers, attention is a powerful tool to capture long-range dependency. However, such power comes at a cost: it incurs a huge computation burden and heavy memory footprint as pairwise token interaction across all spatial locations is computed. A series of works attempt to alleviate this problem by introducing handcrafted and content-agnostic sparsity into attention, such as restricting the attention operation to be inside local windows, axial stripes, or dilated windows. In contrast to these approaches, we propose a novel dynamic sparse attention via bi-level routing to enable a more flexible allocation of computations with content awareness. Specifically, for a query, irrelevant key-value pairs are first filtered out at a coarse region level, and then fine-grained token-to-token attention is applied in the union of remaining candidate regions (\ie, routed regions). We provide a simple yet effective implementation of the proposed bi-level routing attention, which utilizes the sparsity to save both computation and memory while involving only GPU-friendly dense matrix multiplications. Built with the proposed bi-level routing attention, a new general vision transformer, named BiFormer, is then presented. As BiFormer attends to a small subset of relevant tokens in a \textbf{query adaptive} manner without distraction from other irrelevant ones, it enjoys both good performance and high computational efficiency, especially in dense prediction tasks. Empirical results across several computer vision tasks such as image classification, object detection, and semantic segmentation verify the effectiveness of our design. Code is available at \url{https://github.com/rayleizhu/BiFormer}.

Keywords

Cite

@article{arxiv.2303.08810,
  title  = {BiFormer: Vision Transformer with Bi-Level Routing Attention},
  author = {Lei Zhu and Xinjiang Wang and Zhanghan Ke and Wayne Zhang and Rynson Lau},
  journal= {arXiv preprint arXiv:2303.08810},
  year   = {2023}
}

Comments

CVPR 2023 camera-ready

R2 v1 2026-06-28T09:19:02.173Z