MLP-Mixer: An all-MLP Architecture for Vision

TL;DR

This paper introduces MLP-Mixer, a novel vision architecture based solely on multi-layer perceptrons, achieving competitive image classification performance without convolutions or attention mechanisms.

Contribution

The paper demonstrates that pure MLP-based models can match the performance of CNNs and Transformers in vision tasks, challenging existing assumptions about necessary components.

Findings

MLP-Mixer achieves competitive accuracy on image classification benchmarks.

The architecture requires comparable pre-training and inference costs to state-of-the-art models.

Pure MLP models can serve as viable alternatives in computer vision.

Abstract

Convolutional Neural Networks (CNNs) are the go-to model for computer vision. Recently, attention-based networks, such as the Vision Transformer, have also become popular. In this paper we show that while convolutions and attention are both sufficient for good performance, neither of them are necessary. We present MLP-Mixer, an architecture based exclusively on multi-layer perceptrons (MLPs). MLP-Mixer contains two types of layers: one with MLPs applied independently to image patches (i.e. "mixing" the per-location features), and one with MLPs applied across patches (i.e. "mixing" spatial information). When trained on large datasets, or with modern regularization schemes, MLP-Mixer attains competitive scores on image classification benchmarks, with pre-training and inference cost comparable to state-of-the-art models. We hope that these results spark further research beyond the realms of well established CNNs and Transformers.