A Light Weight CNN Based Architecture for the Detection of Early and Late Blight Disease in Tomato Plants in Real-Time Environment

Tomatoes are a globally important vegetable crop, but diseases can have devastating consequences for tomato plants, highlighting the need for prompt identification and treatment of these infections. A variety of machine learning algorithms and Convolutional Neural Network (CNN) models have been proposed in the literature for detecting tomato plant diseases. CNN models leverage the power of deep learning and neural networks. This paper introduces a simplified light weight CNN model comprising ten hidden layers, which surpasses conventional machine learning techniques and pre-trained achieving an impressive accuracy of 99.7 percent. We have collected different types of images samples from Kaggle with the range from 10000-40000 images. The model has been trained and test using different data and the performance was closely examined to make sure the collected data correct. The CNN model after quantization has been implemented through Ryze Tech Tello Mini Drone Quadcopter for the testing real time images.