[ACM MM 2022] SIM-Trans: Structure Information Modeling Transformer for FGVC

Introduction

• In this paper, we propose the structure information modeling transformer (SIM-Trans) that introduces the object structure information into vision transformer for boosting the discriminative feature learning to contain both the appearance and structure information.
• 同时作者也指出了在 FGVC 任务上，ViT 相比 CNN 的优势：“the stacked convolution and pooling operations