Identification of Plant Leaf Disease using Image Augmentation and Deep Learning

Early identification of plant leaf Disease is critical for maintaining agricultural output and food security. Farmers may reduce crop loss and preserve high yields by precisely recognizing illnesses early on and executing prompt preventative steps, such as using pesticides or adopting resistant crop types. Automated disease identification systems offer crop monitoring that is quick and scalable. This study focuses on early stage disease identification using image analysis. This study uses open source the Plant Pathology 2020 - FGVC7 dataset. The dataset consists of more than 7000 images of different plants and categorizes them into Healthy, Scab, Rust, as well as Multiple Disease. The study uses a per-trained Dense-Net and Efficient-Net model for the training and testing of the model. The model uses global average pooling and soft-max activation function due to the multi-classification problem. Dense-Net achieves accuracy for training and testing, 99.87% and 95.26%, while Efficient-Net model archives accuracy for training and testing, 99.22% and 94.54%. However, using DL models such as CNN, and its pre-trained models for plant leaf identification has shown tremendous promise for boosting precision as well as productivity. These algorithms not only give high accuracy rates, but they also pave the way for early identification, individualized treatment strategies, and the development of forecasting techniques to aid illness detection. However, using DL models such as CNN, and its pre-trained models for plant leaf identification has shown tremendous promise for boosting precision as well as productivity. These algorithms not only give high accuracy rates, but they also pave the way for early identification, individualized treatment strategies, and the development of forecasting techniques to aid illness detection.