A New Light Weight Convolutional Neural Network for Mobile Devices

Abstract

In this paper, we propose a light-weight convolutional neural network based on depth-wise separable convolutions and cross stage partial (CSP) network. Dissimilar to MobileNetV3, the proposed network is composed of some CSP blocks to reduce the model size and computational operations. For performance evaluation, Cifar10 and Cifar100 are used for testing. Compared to MobileNetv3, the model size and execution time of the proposed network in PC and mobile device are smaller. Therefore, the experimental results show that the proposed light-weight network can effectively extract visual features for image classification compared with MobileNetV3.