CLOUD-BASED DEEP LEARNING FRAMEWORK FOR AUTOMATED PLANT LEAF DISEASE DETECTION USING IOT AND MULTISPECTRAL IMAGING
DOI: https://doi.org/10.65725/JCISE/1/1/005
JOURNAL OF COMPUTATIONAL INTELLIGENCE SCIENCE AND ENGINEERING (JCISE)
ISSN: 3107-8168
Volume 1 Issue 1, July 2025

Abstract
This study presents an IoT-enabled cloud analytics platform for real-time plant disease detection, integrating field-deployed multispectral sensors with a deep learning pipeline. The system captures leaf images through edge devices, preprocesses them using adaptive histogram equalization, and uploads data to a cloud-based feature extraction module employing a modified GoogleNet architecture. Trained on the Plant Village dataset (54,306 images across 38 disease classes), our model achieves 96.4% classification accuracy with 15% lower computational overhead compared to ResNet-50 benchmarks. The framework implements parallel processing in AWS Lambda, reducing latency to 1.2 seconds per image analysis. Validation on ImageNet-derived agricultural subsets demonstrates robust generalization (F1-score: 0.93) across unseen species. Results indicate a 40% improvement in early disease detection sensitivity over traditional CT imaging approaches, while reducing cloud compute costs by optimizing layer pruning in the feature extraction phase.

Authors: Dr. Kalaivanan E, Mrs. Sharmila K

Keywords: CT imaging; feature extraction; GoogleNet; Computational Cost.