FPGA-Based Plant Identification Through Leaf Veins

Abstract

Identifying and understanding numerous plant species is of utmost importance not only to botanists but also to non-professionals so that they will be able to recognize its name and characteristics. This study is focusing on plant identification with the use of field-programmable gate array (FPGA) device Altera DE1-SoC Cyclone V platform. An openly accessible high-level synthesis (HLS) tool was used to perform image classification of plants leaf using the convolutional neural network (CNN) based on the technology called OpenCL. A total of fifty plants with five leaf samples per plant were being trained using Caffe. The system has yielded an accuracy rate of 88.68% with the aid of a confusion matrix tool.