A machine-learning model for high-resolution computer vision could enable computationally intensive vision applications, such as autonomous driving or medical image segmentation, on edge devices.