PataNET: A Convolutional Neural Networks to Identify Plant from Leaf Images

• Published in: 2019 10th International Conference on Computing, Communication and Networking Technologies (ICCCNT) pp. 1 - 6
• Main Authors: Islam, Md. Majedul, Rabby, AKM Shahahriar Azad, Arfin, Md. Hafizur Rahman, Hossain, Syed Akhter
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.07.2019
• Subjects: Adam optimizer; automatic Learning Rate Reduction; Computer science; Convolution; Convolutional neural network; Convolutional neural networks; Feature extraction; Image recognition; Leaf classification; Training
• DOI: 10.1109/ICCCNT45670.2019.8944667

Plants are everywhere around the Earth. Plant identification is a very important problem for environmental protection and exploration. But getting to know various plants requires taking into account a large number of features which might be easy for a botanist but not easy for ordinary people. So having an automatic Plant identifier which uses a leaf to identifying plants will help many people. Here, a method is proposed where Convolutional Neural Network (CNN) technique is used to classify plants using leaf images. Using Adam optimizer and automatic Learning Rate reduction technique the model gave promising accuracy. This system was trained on 3600 RGB leaf images of 2 categories for 6 different plant species. The model reported promising results with validation accuracy was 95.86% and training accuracy was 96.54%. Different pre-processing techniques such as background whitening, noise removal are used. In the convolutional networks activation function ReLU is used in the hidden layer and Softmax for output layer.