Detecting tomato leaf diseases by image processing through deep convolutional neural networks

Abstract

Machine Learning (ML) and Deep Learning (DL) have already brought unprecedented success in the detection of various diseases of plant leaves, fruits, buds, flowers, etc. Besides, computer science and related field researchers are widely trying to use specific ML and DL methods to classify images and get better results in the field of agriculture and technology. Considering these, Deep Convolutional Neural Networks (DCNN) have been applied in this research. We first applied the Gaussian filter and the Median filter separately on the main dataset and saved the filtered images into two separate directories. We then applied two color models (HSI and CMYK) separately to the images in each directory. Thus, we pre-processed the images in four different ways with the main objective of finding the best combination of the filtering methods and the color models. We then applied our selected DCNN models to each output obtained from the pre-processing steps and finally chose the best methodology based on the accuracy. At last, we have found the highest accuracies (98.27% in Vgg-19, 94.98% in MobileNet-V2, and 99.53% in the ResNet-50) by using the Gaussian Blur and the Gaussian Noise filters with the RGB to CMYK color conversion method.

Keywords

• Leaf disease detection
• Deep learning
• Neural networks
• Color models
• Median filter
• Gaussian filter