Paddy Leaf Disease Detection Using Fine-tuned ConvNext-Tiny Model

Rice is a very important food for people all over the world, and many farmers grow it to earn money and feed their families. But sometimes, rice plants get sick, especially on their leaves. When this happens, the plants don’t grow well and produce less rice. Usually, farmers or experts check the leaves with their eyes to find out if the plant is sick. This takes a lot of time, costs money, and they can sometimes make mistakes. In this project, we created a smart computer system that can look at photos of rice leaves and tell if the plant is healthy or has a disease. We used a special computer model called ConvNeXt, which is very good at understanding pictures. Instead of teaching the model from the beginning, we used a version that had already learned from many images. Then, we trained it again using photos of rice leaves — some healthy and some with diseases. Our dataset had 8 types: one for healthy leaves and seven for different rice leaf diseases. To help the model learn better, we used simple tricks like showing it flipped or rotated pictures (called data augmentation), and dropout, which helps prevent confusion during learning. After training, our model worked very well. It could correctly find diseases 98.44% of the time, and it was very good at being accurate and avoiding mistakes. Farmers can use this system by simply taking a photo of a leaf with their phone, and the system will tell them what disease the plant might have. This saves time, protects the crops, and helps farmers grow better rice. This report explains how we made and tested this helpful tool, step by step. Index Terms-Rice Leaf Disease, Deep Learning, convNext, Image Classification, Transfer Learning, Data Augmentation, Dropout, Precision Agriculture, Smart Farming, Disease Diagnosis, Neural Networks, High Accuracy Detection