What is MobileNet and why is it important for efficient computer vision applications? How does MobileNet reduce computational cost and model size compared to traditional convolutional neural networks? What is the role of depthwise separable convolutions in the MobileNet architecture? How is MobileNet used in mobile and embedded devices for real-time image processing? What are the advantages and limitations of using MobileNet in computer vision tasks?