DecomposeMe: Simplifying ConvNets for End-to-End Learning

TL;DR

DecomposeMe introduces a method using 1D convolutions to simplify ConvNets, reducing parameters and improving performance on large-scale image classification tasks, making them more suitable for embedded devices.

Contribution

The paper presents DecomposeMe, a novel architecture that simplifies ConvNets with 1D convolutions, enabling filter sharing and reducing parameters while maintaining or improving accuracy.

Findings

Achieves 7.7% relative top-1 accuracy improvement on Places2.

Reduces model parameters by 92% compared to VGG-B.

Generalizes to other networks by conversion.

Abstract

Deep learning and convolutional neural networks (ConvNets) have been successfully applied to most relevant tasks in the computer vision community. However, these networks are computationally demanding and not suitable for embedded devices where memory and time consumption are relevant. In this paper, we propose DecomposeMe, a simple but effective technique to learn features using 1D convolutions. The proposed architecture enables both simplicity and filter sharing leading to increased learning capacity. A comprehensive set of large-scale experiments on ImageNet and Places2 demonstrates the ability of our method to improve performance while significantly reducing the number of parameters required. Notably, on Places2, we obtain an improvement in relative top-1 classification accuracy of 7.7\% with an architecture that requires 92% fewer parameters compared to VGG-B. The proposed network is also demonstrated to generalize to other tasks by converting existing networks.