We are delighted to share the publication of “An Edge Computing-Based Solution for Real-Time Leaf Disease Classification Using Thermal Imaging” in the prestigious journal IEEE Geoscience and Remote Sensing Letters [link]. Authored by Públio Elon Correa da Silva and Jurandy Almeida, this paper introduces a groundbreaking approach to integrating edge computing and deep learning for agriculture.
The study focuses on leveraging real-time thermal imaging to classify leaf diseases, a critical step in ensuring crop health and food safety. By evaluating deep learning models such as InceptionV3, MobileNetV1, MobileNetV2, and VGG-16 on resource-constrained devices like the Raspberry Pi 4B, the research demonstrates the feasibility of deploying advanced algorithms outside traditional data centers.
Key findings include:
- Development of a new thermal image dataset for plant disease classification.
- Significant inference time reductions of up to 2.13× with precision reduction techniques, compared to high-end GPUs.
- Effective use of pruning and quantization-aware training to optimize model performance on edge devices.
Presented at SIBGRAPI’24, this study underscores the potential of edge computing to revolutionize precision agriculture. The proposed solution not only reduces dependency on costly hardware but also ensures real-time decision-making capabilities in the field. Congratulations to the authors for their contribution to geoscience, remote sensing, and sustainable agriculture!