Researchers have developed a fully automated AI-IoT system that detects and treats plant diseases in real time using a modified central pivot irrigation setup.
Study: AI-IoT based smart agriculture pivot for plant diseases detection and treatment. Image Credit: i-am-helen/Shutterstock.com
In a recent article published in Scientific Reports, the team introduced this intelligent agriculture system that combines deep learning with IoT infrastructure to streamline crop monitoring and targeted treatment in the field.
Rethinking Traditional Farming
Conventional farming often relies on manual monitoring, which is time-consuming and less efficient, especially when it comes to early detection of plant diseases or timely treatment. In contrast, integrating artificial intelligence (AI) with the Internet of Things (IoT) opens up new possibilities for automating these tasks with precision and speed.
Deep learning models, particularly convolutional neural networks (CNNs), have proven highly effective in classifying plant diseases from images, with models like VGG-16, DenseNet, and ResNet achieving accuracy rates above 97 %. These models are typically trained on extensive datasets such as Plant Village or custom collections of labeled images.
While IoT has already made inroads into agriculture—through sensors, drones, and wireless systems—the challenge has been building a cohesive, autonomous, field-ready system that ties AI-powered decision-making to on-the-ground hardware at scale.
The Current Study
Addressing that gap, the researchers involved with this study designed a multi-layered architecture that blends hardware and software into a seamless, automated solution. At the core is a reengineered central pivot irrigation system embedded with IoT sensors and actuators, all managed by a Raspberry Pi controller.
This setup enables the system to continuously monitor environmental conditions, capture high-resolution images of crop leaves, and act on the data in real time. For disease classification, the team built a robust dataset of around 25,940 images across 11 plant health categories, including both healthy and diseased states, for crops such as tomatoes, peppers, and potatoes.
To improve accuracy and generalization, they applied data augmentation techniques during preprocessing. A pre-trained ResNet50 model, fine-tuned via transfer learning, then classified the images into eight diseased and three healthy categories.
What makes the system field-ready is its tight integration with a mobile application. Farmers use the app to capture leaf images, which are sent to the Raspberry Pi for classification. If disease is detected, the system can automatically trigger treatment through precision sprayers mounted on the irrigation arm. The app also provides real-time updates, diagnoses, and treatment recommendations, making the system both intelligent and user-friendly.
Results and Discussion
The system performed exceptionally well in both controlled environments and field conditions. The ResNet50 model achieved a test accuracy of 99.8 %, with precision, recall, and F1-scores of 100 %, 99.92 %, and 99.91 %, respectively, surpassing many existing models in reliability.
Image augmentation and a well-balanced dataset played a critical role in minimizing overfitting and ensuring consistent performance across various plant types and disease patterns. The mobile app efficiently identified plant health status, enabling fast decision-making, while the hardware proved effective in executing real-time, on-the-spot treatments—reducing the time and labor typically required for manual interventions.
One notable advantage of the system is its integration with a modified pivot irrigation setup, which allows for uniform and efficient treatment distribution, cutting down on both waste and cost. By processing data locally (on the Raspberry Pi) rather than relying on cloud services, the system improves latency, enhances data privacy, and supports broader scalability.
While the study highlights the system's strong potential, it also notes areas for future development. These include testing the platform in more diverse climates and geographies, addressing scalability challenges across larger farm areas, and improving the durability of hardware components under rugged field conditions.
Conclusion
This research introduces a comprehensive AI-IoT platform that pushes precision agriculture closer to full automation. By combining a deep learning-based disease classification system with an upgraded central pivot irrigation setup, the solution offers rapid, accurate detection and treatment of plant diseases. The system’s strong performance in real-world conditions points to its practical benefits—higher yields, lower costs, and more sustainable farming practices.
Looking ahead, future enhancements will focus on broader crop adaptation, refined water and pest management, and better integration across diverse agricultural landscapes.
Journal Reference
Ibrahim A.S., Mohsen S., et al. (2025). AI-IoT based smart agriculture pivot for plant diseases detection and treatment. Scientific Reports 15, 16576. DOI: 10.1038/s41598-025-98454-6, https://www.nature.com/articles/s41598-025-98454-6