Optimum CNN-Based Plant Mutant Classification

Abstract

The study of the observable characteristics of mutants of the same genotype plant interacting with various environmental conditions is important to understand how well the performance of a particular trait in different growth environment. By automating the plant mutant classification process, botanist and agriculture scientist can perform large scale experiments to cultivate plants with useful traits to combat extreme environment conditions. This research aims to construct an optimum convolutional neural network (CNN) for image-based plant mutant classification task. Optimum parameters for 1) number of convolutional layers, 2) number of neurons in fully connected (FC) layer and 3) Number of FC layers are found in this paper. The possibility to improve success classification rate was explored by applying image pre-processing methods. Experimental results show that under optimum condition, CNN classification system without pre-processing algorithm shows the best success recognition rate of 97.90%.