TY - JOUR
T1 - Tomato plant leaf disease identification and classification using deep learning
AU - Tahir, Hannia
AU - Samimi, Parnia
PY - 2023/9/23
Y1 - 2023/9/23
N2 - Crops are an integral part of the agricultural industry in South Asian countries as they have fertile soil and advantageous weather conditions. How-ever, the crops suffer from various plant diseases that influence the quality of the crops, which can be due to the lack of proper diagnosis and/or inefficient and untimely diagnoses of crop diseases. Diagnostic approaches of crop diseases are varied. Some can be diagnosed grossly, and some need especial laboratory ef-forts. Gross diagnosis approach is quite inaccurate and sometimes tricky. On the other hand, laboratory diagnosis is a time-consuming process and quite costly. In this project we developed an automated image processing method to enhance efficiency and accuracy in the gross assessment. This process accelerates the diag-nosis process and can be considered as a preliminary method for diagnosis. We utilized deep learning to detect diseases in tomato crops using leaf images. The process involves building a convolutional neural network using a pre-trained VGG16 model that pre-processes the images according to its requirements and performs segmentation on images before training and testing the data. The model obtained an accuracy of 95%, taking only 30 minutes to run, train, test, and classify the 18,160 images to their respective 10 classes thus being very time-efficient and did not need a laptop with higher power processor, which makes the model accessible for other devices. Due to the low complexity of the model, it can be implemented on smaller devices without the need of a fast processor.
AB - Crops are an integral part of the agricultural industry in South Asian countries as they have fertile soil and advantageous weather conditions. How-ever, the crops suffer from various plant diseases that influence the quality of the crops, which can be due to the lack of proper diagnosis and/or inefficient and untimely diagnoses of crop diseases. Diagnostic approaches of crop diseases are varied. Some can be diagnosed grossly, and some need especial laboratory ef-forts. Gross diagnosis approach is quite inaccurate and sometimes tricky. On the other hand, laboratory diagnosis is a time-consuming process and quite costly. In this project we developed an automated image processing method to enhance efficiency and accuracy in the gross assessment. This process accelerates the diag-nosis process and can be considered as a preliminary method for diagnosis. We utilized deep learning to detect diseases in tomato crops using leaf images. The process involves building a convolutional neural network using a pre-trained VGG16 model that pre-processes the images according to its requirements and performs segmentation on images before training and testing the data. The model obtained an accuracy of 95%, taking only 30 minutes to run, train, test, and classify the 18,160 images to their respective 10 classes thus being very time-efficient and did not need a laptop with higher power processor, which makes the model accessible for other devices. Due to the low complexity of the model, it can be implemented on smaller devices without the need of a fast processor.
KW - VGG16
KW - Plant Disease Detection
KW - Deep Learning
KW - Tomato Plant
UR - https://www.open-access.bcu.ac.uk/13675/
M3 - Article
SN - 2367-4520
JO - ICACIn 2022
JF - ICACIn 2022
ER -