MLP-Mixer: An all-MLP Architecture for Vision
It challenges the necessity of convolutions and attention in vision models, potentially sparking research beyond CNNs and Transformers.
The paper tackles the problem of designing neural network architectures for computer vision by proposing MLP-Mixer, an all-MLP model that achieves competitive scores on image classification benchmarks, with pre-training and inference costs comparable to state-of-the-art models.
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.