CycleMLP: A MLP-like Architecture for Dense Prediction

TL;DR

CycleMLP introduces a versatile MLP-like architecture that handles various image sizes efficiently, surpasses existing models in dense prediction tasks, and maintains low computational complexity, making it suitable for object detection and segmentation.

Contribution

The paper proposes CycleMLP, an MLP-like architecture with linear complexity and adaptability to different image sizes, outperforming existing MLPs and Transformer models in dense visual prediction tasks.

Findings

CycleMLP outperforms Swin-Tiny by 1.3% mIoU on ADE20K.

CycleMLP achieves competitive results in object detection and segmentation.

CycleMLP demonstrates strong zero-shot robustness on ImageNet-C.

Abstract

This paper presents a simple MLP-like architecture, CycleMLP, which is a versatile backbone for visual recognition and dense predictions. As compared to modern MLP architectures, e.g., MLP-Mixer, ResMLP, and gMLP, whose architectures are correlated to image size and thus are infeasible in object detection and segmentation, CycleMLP has two advantages compared to modern approaches. (1) It can cope with various image sizes. (2) It achieves linear computational complexity to image size by using local windows. In contrast, previous MLPs have $O(N^2)$ computations due to fully spatial connections. We build a family of models which surpass existing MLPs and even state-of-the-art Transformer-based models, e.g., Swin Transformer, while using fewer parameters and FLOPs. We expand the MLP-like models' applicability, making them a versatile backbone for dense prediction tasks. CycleMLP achieves competitive results on object detection, instance segmentation, and semantic segmentation. In particular, CycleMLP-Tiny outperforms Swin-Tiny by 1.3% mIoU on ADE20K dataset with fewer FLOPs. Moreover, CycleMLP also shows excellent zero-shot robustness on ImageNet-C dataset. Code is available at https://github.com/ShoufaChen/CycleMLP.