High resource consumption and latency in deep learning models prevent deployment on edge devices. Reducing complexity enables real-time processing and lower power consumption.