Disease Detection with Hybrid Architecture of ResNet and VGG16 Networks in Real Time

Modern agriculture faces a critical challenge in ensuring the health and yield of crops, with the early detection and treatment of diseases playing a pivotal role. Traditional methods reliant on manual inspection are not only time-consuming but also prone to errors, leading to significant crop losses and economic impact. The importance of this research lies in mitigating these inefficiencies by developing a robust ML model capable of accurately identifying diseased tomato leaves in the plants. Datasets considered while algorithm training comprise around 6 thousand to 8 thousand images, including both healthy and diseased leaves. The primary gap this research addresses is the inadequacy of conventional methods in providing timely and accurate disease detection. The research aims to harness the power of advanced ML algorithms including Visual Geometry Group 16 & Residual Network for the deep feature extraction. The model is trained, validated, and tested using a split dataset approach to ensure high accuracy and reliability. Key findings indicate that the model achieves an accuracy of around 95% in simulations, although a slight drop in accuracy is observed in real-time applications. Despite this, the ML-based approach significantly surpasses traditional methods, offering a more efficient and scalable solution to find the diseased leaves soon in tomato. These findings highlight potential in ML to transform agricultural practices by providing timely and accurate disease detection, reducing the reliance on manual labor, and contributing to increased crop yields and sustainable farming practices.