Plant Disease Detection Using Machine Learning and Image Segmentation Techniques

Abstract

Crop infections are more frequent now, which causes farmers to suffer large losses every year. The traditional approach of diagnosing plant diseases by visual inspection takes a lot of time, and the pathologist's knowledge has a significant impact on how accurately the illness is found. With the help of the recommended technique, farmers will have a tool for quickly and accurately diagnosing plant illnesses, which will save them time and money. The approach uses the pre-trained EfficientNetB3 model, which was trained on a sizable dataset of photographs. It is based on transfer learning. The collection for this project includes images of 14 different types of plant leaf diseases. Training, validation, and testing data sets are created after pre-processing, supplementing, and dividing the data. The pre-trained EfficientNetB3 model is utilised to generate a CNN model using TensorFlow, which is then used to train the CNN model. The model is evaluated based on several performance metrics, and the results show that it is quite effective in identifying plant diseases.With the help of the recommended technique, farmers will have a tool for quickly and accurately diagnosing plant illnesses, which will save them time and money. The approach uses the pre-trained EfficientNetB3 model, which was trained on a sizable dataset of photographs. It is based on transfer learning. The collection for this project includes images of 14 different types of plant leaf diseases. Training, validation, and testing data sets are created after pre-processing, supplementing, and dividing the data. A CNN model is built using TensorFlow and trained using an EfficientNetB3 model that has already been trained. The model is evaluated based on many performance metrics, and the results show that it has a high level of diagnostic precision for plant diseases.