Abstract: In recent years, considerable research has focused on the use of custom hardware to accelerate deep learning on edge devices. However, the end-to-end flow of deep learning includes ...