Quantized alpha–WIoU YOLOv5 algorithm for rice leaf disease detection

This paper presents an innovative method based on convolutional neural networks for early detection and diagnosis of diseases in rice plants. The YOLOv5n model is modified by replacing its original loss function with the alpha-WIoU function to enhance feature learning and increase accuracy. The model is then quantized to reduce its size and optimize it for low-end devices. Our improved model achieves an accuracy of 93%, which is 2.6% higher than the original YOLOv5n, while reducing the model size by 45%. Additionally, it processes each image in just 0.554 seconds on a CPU, nearly three times faster than the original model. The results show that the proposed model is more accurate, lighter and faster, suitable for practical implementation, contributing to the development of smart agriculture