ConvMLP: Hierarchical convolutional MLPS for vision

In the past decade, deep learning has made breakthroughs in most computer vision tasks, and most solutions are based on deep convolution neural networks. Recently, MLP-based architectures, which consist of a sequence of consecutive multi-layer perceptron blocks, have been found to achieve results comparable to those of convolutional and transformer-based methods. However, most MLP-based architectures adopt spatial MLPs which take fixed dimension inputs, therefore making it difficult to apply them to downstream computer vision tasks, such as object detection and semantic segmentation. Moreover, single-stage designs further limit performance in other computer vision tasks and fully connected layers bear heavy computation. To tackle these problems, we propose ConvMLP: a hierarchical Convolutional MLP for visual recognition, which is a lightweight, stage-wise, co-design of convolution layers and MLPs. In particular, ConvMLP-S achieves 76.8\% top-1 accuracy on ImageNet-1k with 9M parameters and 2.4 GMACs (15\% and 19\% of MLP-Mixer-B/16, respectively). Experiments on object detection and semantic segmentation further show that visual representation learned by ConvMLP can be seamlessly transferred and achieve competitive results with fewer parameters.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Electrical & Computer Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2022

Author dc:creator

• Li, Jiachen

Contributors dc:contributor

• Shi, Humphrey

• MLP
• Vision

Statement dc:rights

• Copyright 2022 Jiachen Li

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/116263

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F116263