Bangladeshi Plant Leaf Classification and Recognition Using YOLO Neural Network

Abstract

Now a days, recognition of plant species and leaves holds a great importance in the field of medical science, environmental issues like maintaining ecological balance, preserving distinct plants etc. Using Deep Convolutional Neural Network (CNN) as a classifier, has shown tremendous success on the field of classification and detection task. In this paper, we have proposed to use the YOLOv2 model as a classifier through which we have trained the model using our leaf dataset. In this transfer learning approach our target leaf data set was used to classify leaves as our target task. Both recognition and localization of the leaves are done through this work. Multiple leaves detection and classification is also an achievement of this work. Approximately 96% classification accuracy was achieved to classify the leaves which had also shown a satisfactory localization accuracy also. Both recognition and localization of leaves will bring a success to the researchers in the field of botany, medicinal plant analysis also.