MobileNets are efficient convolutional neural networks designed for mobile and embedded vision applications. They use depthwise separable convolutions to create lightweight models that balance latency and accuracy. Two hyperparameters, width multiplier and resolution multiplier, allow model builders to adjust model size and performance based on application needs. MobileNets have been tested on various tasks, including image classification, object detection, fine-grained classification, face attributes, and large-scale geolocation. They outperform other models in terms of size, speed, and accuracy. MobileNets are also effective in applications like face recognition and geolocation, where they achieve high performance with reduced computational requirements. The paper also discusses the use of MobileNets in object detection and face embeddings, showing their versatility and efficiency. Overall, MobileNets provide a compact and efficient solution for mobile vision tasks.MobileNets are efficient convolutional neural networks designed for mobile and embedded vision applications. They use depthwise separable convolutions to create lightweight models that balance latency and accuracy. Two hyperparameters, width multiplier and resolution multiplier, allow model builders to adjust model size and performance based on application needs. MobileNets have been tested on various tasks, including image classification, object detection, fine-grained classification, face attributes, and large-scale geolocation. They outperform other models in terms of size, speed, and accuracy. MobileNets are also effective in applications like face recognition and geolocation, where they achieve high performance with reduced computational requirements. The paper also discusses the use of MobileNets in object detection and face embeddings, showing their versatility and efficiency. Overall, MobileNets provide a compact and efficient solution for mobile vision tasks.