EViT: An Eagle Vision Transformer with Bi-Fovea Self-Attention

arXiv:2310.06629v3 Announce Type: replace
Abstract: Thanks to the advancement of deep learning technology, vision transformers has demonstrated competitive performance in various computer vision tasks. Unfortunately, vision transformers still faces some challenges such as high computational complexity and absence of desirable inductive bias. To alleviate these issues, we propose a novel Bi-Fovea Self-Attention (BFSA) inspired by the physiological structure and visual properties of eagle eyes. This BFSA is used to simulate the shallow and deep fovea of eagle vision, prompting the …

arxiv attention cs.cv self-attention transformer type vision