Alpha-EIOU-YOLOv8 is an improved algorithm for rice leaf disease detection. This paper presents a modified YOLOv8 model that replaces the original Box Loss function with a combination of EIoU loss and α-IoU loss to enhance the performance of rice leaf disease detection. A two-stage approach is proposed to achieve high accuracy in rice leaf disease identification using AI algorithms. In the first stage, images of rice leaf diseases in the field are automatically collected and separated into blast leaf, leaf folder, and brown spot sets. In the second stage, the YOLOv8 model is trained on the proposed image dataset and deployed on IoT devices to detect and identify rice leaf diseases. The experimental results show that the accuracy of the proposed model reaches up to 89.9% on a dataset of 3175 images, outperforming existing methods. The model achieves higher accuracy than YOLOv7 and YOLOv5. The proposed method also demonstrates better performance in terms of precision, recall, and F1-score. The model is implemented on a Raspberry Pi 4 Model B, a low-cost device, and is capable of providing timely disease detection in rice fields. The research also provides a new rice leaf disease image dataset of the three most common diseases in Vietnam. The results show that the proposed method is effective for rice leaf disease detection and has potential for practical application in agriculture.Alpha-EIOU-YOLOv8 is an improved algorithm for rice leaf disease detection. This paper presents a modified YOLOv8 model that replaces the original Box Loss function with a combination of EIoU loss and α-IoU loss to enhance the performance of rice leaf disease detection. A two-stage approach is proposed to achieve high accuracy in rice leaf disease identification using AI algorithms. In the first stage, images of rice leaf diseases in the field are automatically collected and separated into blast leaf, leaf folder, and brown spot sets. In the second stage, the YOLOv8 model is trained on the proposed image dataset and deployed on IoT devices to detect and identify rice leaf diseases. The experimental results show that the accuracy of the proposed model reaches up to 89.9% on a dataset of 3175 images, outperforming existing methods. The model achieves higher accuracy than YOLOv7 and YOLOv5. The proposed method also demonstrates better performance in terms of precision, recall, and F1-score. The model is implemented on a Raspberry Pi 4 Model B, a low-cost device, and is capable of providing timely disease detection in rice fields. The research also provides a new rice leaf disease image dataset of the three most common diseases in Vietnam. The results show that the proposed method is effective for rice leaf disease detection and has potential for practical application in agriculture.