Automated Identification of Crop Disease Using Deep Neural Network with Proximal Policy Optimization Technique

Detection of crop diseases at an early stage with good accuracy and efficiency is crucial for gaining maximum yields. The benefits that precision agriculture yields can save the losses it incurs. Routine techniques, like manual identification and some traditional machine learning algorithms of plant disease diagnosis are prone to human error, time-consuming, and labor intensive. This research proposes an efficient automated crop disease detection system using computer vision that involves reinforcement learning techniques. This work utilizes the deep learning ResNet-9 model to captures crop features efficiently and classifies with reinforcement learning based Proximal Policy Optimization (PPO), thus making the decision optimal. The proposed ResNet-9 Model trained on a diverse crop image dataset of healthy and diseased plant leaves from Kaggle data source, which contains Apple, Corn, Potato, and Tomato plant leaves. The ResNet-9 model obtained excellent accuracy of 93.94% in automated feature extraction and classification, at whereas the PPO reinforcement learning classifier further enhances the predicting power of the system with a classification accuracy about 98.74%. All the above techniques couple disease identification and achieve the same in an efficient manner related to classification for proactive management strategies in agricultural practices. The proposed system leverages the strength of reinforcement learning, promises to revolutionize disease diagnosis in agriculture. This research forms a very promising framework for scalable implementation extended to real-time precision farming applications.