FFT-Based Deep Learning Deployment in Embedded Systems

The excellence of deep learning has also resulted in explorations of several emerging real-world applications, such as self-driving systems [3], automatic machine translations [4], drug discovery and toxicology [5]. The deep learning is based on the structure of deep neural networks (DNNs), which consist of multiple layers of various types and hundreds to thousands of neurons in each layer. Recent evidence has revealed that the network depth is of crucial importance to the success of deep learning, and many deep learning models for the challenging ImageNet dataset are sixteen to thirty layers deep [1]. Deep learning achieves significant improvement in overall accuracy by extracting complex and high-level features at the cost of considerable up-scaling in the model size. In the big data era and driven by the development of semiconductor technology, embedded systems are now becoming an essential computing platform with ever-increasing functionalities. At the same time, researchers around the world from both academia and industry have devoted significant efforts and resources to investigate, improve, and promote the applications of deep learning in embedded systems [6].