MUMBAI, India, Nov. 2 -- Intellectual Property India has published a patent application (202641124479 A) filed by St. Joseph'S Institute Of Technology; V. Nithin; M. Nataraj; and G. Mahalakshmi on October 17, 2026, for A Multimodal Deep Vision-Tabular Fusion Approach For Real-Time Sugarcane Disease And Pest Management.
Inventors include St. Joseph'S Institute Of Technology; V. Nithin; M. Nataraj; and G. Mahalakshmi.
The application for the patent was published on October 30, 2026, under issue no. 44/2026.
Abstract: Early recognition of sugarcane diseases and insect pests is vital to reduce crop losses and ensure field management in time. Although deep learning methods have significantly contributed to the automation of crop diagnosis, existing methods often depend only on visual inputs and show poor performance due to low robustness to difficult farming environments. Besides, high computation requirements could impede applying these models on cost-effective edge devices commonly used in the agricultural domain. In order to overcome those shortcomings, we present a multimodal crop diagnosis framework that combines analysis of visual inputs and context- aware agriculture knowledge. We use YOLOv8s for detecting insect pests, YOLOv8sseg for segmentation and quantification of disease severity, and TabNet for processing of agricultural context parameters and farmers’ observations. Joint learning of both types of inputs allows improving the ability of distinguishing between disease symptoms and abiotic stresses. For more practical use in agriculture settings, the framework provides weather information offline, multilingual support, and text-to-speech capabilities. On an independent test dataset, our approach achieved mAP@0.50 of 0.891 and mAP@0.50:0.95 of 0.763. Full framework operated at 38 FPS on NVIDIA RTX 3080 GPU and could be deployed on Raspberry Pi 4B for periodic monitoring of crops at edge.
Disclaimer: Curated by HT Syndication.
