MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications

• DOI: 10.48550/arxiv.1704.04861
• Publication Date: 2017-01-01
• Authors: Andrew Howard; Menglong Zhu; Bo Chen; Dmitry Kalenichenko; Weijun Wang; Tobias Weyand; Marco Andreetto; Hartwig Adam
• Keywords: Deep Neural Networks

We present a class of efficient models called MobileNets for mobile and embedded vision applications. are based on streamlined architecture that uses depth-wise separable convolutions to build light weight deep neural networks. introduce two simple global hyper-parameters efficiently trade off between latency accuracy. These allow the model builder choose right sized their application constraints problem. extensive experiments resource accuracy tradeoffs show strong performance compared other popular ImageNet classification. then demonstrate effectiveness across wide range applications use cases including object detection, finegrain classification, face attributes large scale geo-localization.