Work

Deep Learning and Edge Computing in Agriculture: A Comprehensive Review of Recent Trends and Innovations

Year: 2025

Type: article

Abstract: Rice is a vital staple for over half of the global population, yet its production is significantly threatened by leaf diseases caused by fungal and bacterial pathogens. Early and accurate detection of such diseases is critical to minimizing crop loss, particularly under conditions of labor shortages and climate variability. Traditional inspection methods are labor-intensive and error-prone, highlighting the need for automated, intelligent solutions. This systematic literature review explores recent advancements in deep learning and edge computing for rice leaf disease detection, focusing on challenges such as small and overlapping lesion detection and real-time model deployment. These lesions are difficult to detect due to their limited size, irregular shapes, and visual similarity to healthy areas, especially in field conditions. A comprehensive search across five databases (2020–2024) using the keywords “rice leaf disease detection,” “object detection,” and “edge computing” identified 2,072 articles. After applying inclusion criteria, 76 peer-reviewed journal articles were selected. Among the models reviewed, You Only Look Once version 7 (YOLOv7) achieved the highest accuracy (99% mAP), outperforming Faster Region-Based Convolutional Neural Network (Faster R-CNN) and YOLOv8. Edge devices like NVIDIA Jetson Nano demonstrated real-time capabilities, achieving 0.0549 seconds latency per image using YOLOv7-tiny at 640×640 resolution. Vision Transformers (ViTs) also performed well (up to 98% mAP) but faced challenges related to computational and data efficiency. Future directions include leveraging lightweight models, solar-powered edge devices, and hybrid ViT–CNN architectures. This review underscores the transformative role of AI technologies in precision agriculture, offering scalable tools to improve crop monitoring and global food security.

Source: IEEE Access

DOI: 10.1109/access.2025.3595621