AUTHOR=Shoaib Muhammad , Hussain Tariq , Shah Babar , Ullah Ihsan , Shah Sayyed Mudassar , Ali Farman , Park Sang Hyun 

TITLE=Deep learning-based segmentation and classification of leaf images for detection of tomato plant disease

JOURNAL=Frontiers in Plant Science

VOLUME=Volume 13 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2022.1031748

DOI=10.3389/fpls.2022.1031748

ISSN=1664-462X

ABSTRACT=Plants contribute significantly to the global food supply. Various Plant diseases can result in production losses, which can be avoided by maintaining vigilance. However, monitoring plant diseases manually is time-consuming and error prone. To reduce the likelihood of disease severity, we have two options. One technique is the use of machine vision. In the alternative method, the severity of the disease can be diminished through the use of a computer and the cooperation of humans (i.e., artificial intelligence). These methods can also eliminate the disadvantages of manual observation. In this work, we proposed using it to diagnose tomato diseases. We utilized an architecture for deep learning based on a recently developed convolutional neural network dubbed Inception Neural Network and 18,161 simple and segmented tomato leaf images. Using supervised learning, Inception Net, an updated deep convolutional neural network model, recognizes tomato diseases in this paper. The model is trained on over 1800 images, some of which are plain while the remainder are segmented leaf images to segment the region of interest, U-net and Modified U-net, two cutting-edge segmentation models, are utilized (leaves). There is also an examination of the presentation of binary arrangement (healthy and diseased leaves), six-level classification (healthy and other ailing leaf groups), and ten-level classification (healthy and other types of ailing leaves) models. The Modified U-net segmentation model outperforms the simple U-net segmentation model by 98.66 percent, 98.5 IoU score, and 98.73 percent on the dice. InceptionNet1 achieves 99.95% accuracy for binary classification problems and 99.12% accuracy for classifying six segmented class images; InceptionNet outperformed the Modified U-net model to achieve higher accuracy. In conclusion, for images of segmented leaves from ten distinct classes, InceptionNet3 achieved 99.89 percent accuracy. Using the experimental results, it is possible to conclude that a network with a greater number of layers can classify segmented images better than plain images. The experimental results of our proposed method for classifying plant diseases demonstrate that it outperforms the methods currently available in the literature.