面向移动端的轻量化卷积神经网络结构

Lightweight convolutional neural network structure for mobile

随着移动设备的大量普及,将卷积神经网络应用于移动设备具有极大的实用价值。虽然随着技术的发展,目前移动设备的计算能力和存储资源都有了极大的提高,但是在移动设备上运行卷积神经网络仍然具有很大的挑战。为了解决这个问题,提出了一种轻量化的卷积神经网络结构S-MobileNet。该结构可以显著地减少网络模型的参数量以及降低模型的计算复杂度。为了全面评测S-MobileNet的性能,在CIFAR-10、CIFAR-100和ImageNet等图像分类数据集上进行了相关实验。实验结果表明,所提出的网络结构在保持MobileNetV2同等准确度的前提下,网络模型的参数量较MobileNetV2减少了近1/3,计算复杂度较MobileNetV2降低了近40%。

With the proliferation of mobile devices,the application of convolutional neural networks to mobile devices has great practical value.Although the computing power and storage resources of mobile devices have been greatly improved with the development of technology,running convolutional neural networks on mobile devices still has great challenges.To solve this problem,we proposed a lightweight convolutional neural network structure named S-MobileNet.This network structure can significantly reduce the number of parameters of the network model and reduce the computational complexity of the model.In order to comprehensively evaluate the performance of S-MobileNet,we conducted experiments on image classification datasets such as CIFAR-10,CIFAR-100 and ImageNet.Experimental results show that while the proposed network structure maintains the same accuracy as MobileNetV2,the parameter size is reduced by nearly 1/3,and the computational complexity is reduced by nearly 40%.