An Enhanced Hybrid MobileNet

TL;DR

This paper introduces an improved MobileNet architecture that reduces computational cost and enhances accuracy by replacing the resolution multiplier with a depth multiplier and integrating advanced pooling methods, validated on CIFAR.

Contribution

The paper proposes a novel architecture that replaces the resolution multiplier with a depth multiplier and combines it with Fractional Max Pooling or max pooling to improve MobileNet.

Findings

Reduces computational cost on CIFAR dataset

Increases accuracy compared to original MobileNet

Effective pooling strategies enhance model performance

Abstract

Complicated and deep neural network models can achieve high accuracy for image recognition. However, they require a huge amount of computations and model parameters, which are not suitable for mobile and embedded devices. Therefore, MobileNet was proposed, which can reduce the number of parameters and computational cost dramatically. The main idea of MobileNet is to use a depthwise separable convolution. Two hyper-parameters, a width multiplier and a resolution multiplier are used to the trade-off between the accuracy and the latency. In this paper, we propose a new architecture to improve the MobileNet. Instead of using the resolution multiplier, we use a depth multiplier and combine with either Fractional Max Pooling or the max pooling. Experimental results on CIFAR database show that the proposed architecture can reduce the amount of computational cost and increase the accuracy simultaneously.