Edge-AI for Agriculture: Lightweight Vision Models for Disease Detection in Resource-Limited Settings
Harsh Joshi
TL;DR
The paper tackles disease detection in orange crops under resource limitations by building a lightweight, edge-optimized Vision pipeline that unifies detection, species classification, and disease segmentation. It evaluates multiple state-of-the-art architectures (e.g., YOLOv8-S for detection, Vision Transformer and MobileNetV3 for classification, LinkNet and FPN for segmentation) on a compact, market-collected dataset spanning five orange species and five diseases, with 5-fold cross-validation and ImageNet pretraining. Key findings show that ViT achieves 96% accuracy in species classification (high but with a large model size), while YOLOv8-S delivers rapid detection (mAP50 0.949, ~10.9 ms per image), and LinkNet offers solid segmentation (IoU ~0.904) with moderate sizes. The study demonstrates that effective edge AI for agricultural tasks is feasible with limited data (about 50 images per class), enabling on-device, real-time disease monitoring and offering a path toward data-efficient, federated, and compressed models for broader deployment.
Abstract
This research paper presents the development of a lightweight and efficient computer vision pipeline aimed at assisting farmers in detecting orange diseases using minimal resources. The proposed system integrates advanced object detection, classification, and segmentation models, optimized for deployment on edge devices, ensuring functionality in resource-limited environments. The study evaluates the performance of various state-of-the-art models, focusing on their accuracy, computational efficiency, and generalization capabilities. Notable findings include the Vision Transformer achieving 96 accuracy in orange species classification and the lightweight YOLOv8-S model demonstrating exceptional object detection performance with minimal computational overhead. The research highlights the potential of modern deep learning architectures to address critical agricultural challenges, emphasizing the importance of model complexity versus practical utility. Future work will explore expanding datasets, model compression techniques, and federated learning to enhance the applicability of these systems in diverse agricultural contexts, ultimately contributing to more sustainable farming practices.