The paper introduces MobileNets, a class of efficient models designed for mobile and embedded vision applications. MobileNets are based on a streamlined architecture that uses depthwise separable convolutions to build lightweight deep neural networks. The authors introduce two hyperparameters—width multiplier and resolution multiplier—to allow model builders to trade off between latency and accuracy, making it easier to choose the right-sized model for specific applications. Extensive experiments on ImageNet show strong performance compared to other popular models. MobileNets are then demonstrated across various applications, including object detection, fine-grained classification, face attributes, and large-scale geo-localization. The paper also discusses prior work in building small and efficient neural networks and provides a detailed description of the MobileNet architecture, including the core layers and training methods. The results show that MobileNets can achieve high accuracy while being significantly smaller and faster than existing models.The paper introduces MobileNets, a class of efficient models designed for mobile and embedded vision applications. MobileNets are based on a streamlined architecture that uses depthwise separable convolutions to build lightweight deep neural networks. The authors introduce two hyperparameters—width multiplier and resolution multiplier—to allow model builders to trade off between latency and accuracy, making it easier to choose the right-sized model for specific applications. Extensive experiments on ImageNet show strong performance compared to other popular models. MobileNets are then demonstrated across various applications, including object detection, fine-grained classification, face attributes, and large-scale geo-localization. The paper also discusses prior work in building small and efficient neural networks and provides a detailed description of the MobileNet architecture, including the core layers and training methods. The results show that MobileNets can achieve high accuracy while being significantly smaller and faster than existing models.