Leaf Disease Detection Using Machine Learning in Conjunction with Image Processing

Convolutional Neural Networks (CNNs) are a popular and useful deep learning application in agriculture for the identification of plant diseases. Given that CNNs excel at image-based tasks, they may be used effectively to recognise plant illnesses from photographs involving their leaves or additional pertinent elements. Take pictures of both healthy and diseased plants to create a dataset of photographs. By photographing the plants in the agricultural area, it can develop its databases or locate publically accessible ones. For this, the collection contains mainly Guava, Jackfruit, and Mango leaves. To maintain consistency, resize all photos to the same dimension (for example, 224 × 224 pixels). Create validation, training, and evaluation sets from the dataset (for example, a 70–15-15 split). Normalise the picture pixel values to lie anywhere from zero to one. Create the architecture for CNN. Models that have been trained like VGG16, ResNet, or Inception are frequently used as a starting point and then fine-tuned for your particular application. If have a big enough dataset, you may even create your own CNN architecture. In this study, the PlantVillage dataset and self dataset were utilised. Using offline supplementation from the initial set of data, this information set has been reconstructed. This collection includes thirty-three distinct classes and around 76,000,000 RGB photos of both healthy and sick crop leaves.