Case study of convolutional neural network implemented on FPGA

Convolutional neural network (CNN) has become a key research area in artificial intelligent technology. A case study of garbage identification and classification using convolutional neural network is presented in this paper. The CNN model with five convolutional layers is used to extract feature maps from garbage pictures, and classifies six types of domestic garbage. ZYNQ-7020 is employed to deploy the CNN model in order to take advantage of high efficiency, low power consumption and flexibility of FPGA platform. Additionally, the classification results are obtained at the FPGA terminal by running Python code through PYNQ. The study demonstrates an efficient approach to practically deploy CNN model and implement image classification on FPGA platform.