Performance Analysis of Different Deep Learning Algorithms for Rice Leaf Disease Detection

Rice is a staple food in Bangladesh, but various bacterial, viral, or fungal diseases can significantly reduce rice production and cause economic losses. Accurate and timely detection of these diseases is crucial for farmers to take appropriate measures. However, current methods for recognizing rice leaf diseases are limited by variations in image backgrounds and capture conditions. To address this challenge, we propose a unique method for identifying rice diseases based on deep convolutional neural networks (CNNs), including VGG19, ResNet50, and InceptionV3. We trained the models to recognize three typical rice diseases using real photographs of damaged and healthy rice stems and leaves taken in an experimental rice field. VGG19 was applied in 19 layers, while InceptionV3 helped achieve higher accuracy. The CNN model extracted features and classified images with greater accuracy using layers. ResNet50 was used to extract more key features from the images. Our simulation results show that the proposed technique is efficient and can effectively identify rice diseases. Deep learning-based automatic detection and diagnosis of rice diseases can contribute to addressing the global rice demand challenge and reduce economic losses for farmers.