FViT: A Focal Vision Transformer with Gabor Filter

arXiv:2402.11303v1 Announce Type: new
Abstract: Vision transformers have achieved encouraging progress in various computer vision tasks. A common belief is that this is attributed to the competence of self-attention in modeling the global dependencies among feature tokens. Unfortunately, self-attention still faces some challenges in dense prediction tasks, such as the high computational complexity and absence of desirable inductive bias. To address these above issues, we revisit the potential benefits of integrating vision transformer with Gabor filter, and propose a Learnable …

abstract arxiv attention belief challenges computational computer computer vision cs.cv dependencies feature filter global modeling prediction progress self-attention tasks tokens transformer transformers type vision vision transformers