MUMBAI, India, Aug. 10 -- Intellectual Property India has published a patent application (202611074480 A) filed by Manipal University Jaipur on June 16, 2026, for An Explainable Deep Learning System For Plant Disease Diagnosis Using Efficientnet-B3 And Gradcam.
Inventors include Ayush Anand; and Nandini Babbar.
The application for the patent was published on July 31, 2026, under issue no. 31/2026.
Abstract: The present invention relates to an explainable plant disease diagnosis system for automated detection of plant diseases from leaf images. The system comprises a pre- processing module for receiving and normalizing plant leaf images, a disease classification engine based on an EfficientNet-B3 neural network trained through a two- stage training process, an explainability module generating GradCAM heatmaps, an uncertainty assessment module for identifying low-confidence predictions, a reporting module generating diagnostic outputs and reports, and a storage module for maintaining prediction records. The disease classification engine utilizes AdamW optimization, label smoothing, and adaptive learning rate reduction. The system supports multiple crop species and disease classes using a single model. The system processes individual or batch images and automatically highlights infected regions responsible for a prediction. Diagnostic outputs include disease predictions, confidence scores, uncertainty indications, and visual explanations to improve reliability and usability.
Disclaimer: Curated by HT Syndication.
