Processing large-scale remote sensing imagery requires excessive time and memory resources. Reducing these bottlenecks allows for real-time object recognition and segmentation.