On the Use of FPGAs to Implement CNNs: A Brief Review

Convolutional Neural Network (CNN) is a subclass of deep neural network that has gained popularity in recent years. CNN has revolutionized the execution of tasks such as natural language processing, image classification, and voice recognition. However, the performance of CNNs is often limited by the hardware available for training large sets of data. Graphical Processing Units (GPUs) have been shown to achieve good performance with CNN-based applications, however, GPU is expensive and is not suitable for all applications. In recent years, and for various reasons researchers have shifted their focus to Field Programmable Gate Arrays (FPGAs) and even other edge devices like microcontrollers to execute CNN models. This paper provides a survey of a number of applications where FPGAs are used in the implementation of various CNN-based models. The survey provides the reader with a compact and informative insight into recent efforts in this domain.