MUMBAI, India, Oct. 26 -- Intellectual Property India has published a patent application (202611108087 A) filed by Chitkara University on September 08, 2026, for A Confidence-Aware Multitask Deep-Learning System And Method For Plant Disease And Nutrient- Deficiency Detection On A Resource-Constrained Edge Device.
Inventor includes Dr. Raj Gaurang Tiwari.
The application for the patent was published on October 23, 2026, under issue no. 43/2026.
Abstract: A confidence-aware multitask deep-learning system for plant disease and nutrient-deficiency detection on an edge device is disclosed, comprising an image acquisition unit, an image quality module, a resource-monitoring module, an adaptive model-selection module, and a multitask deep learning module having a shared feature extractor coupled to a disease head, a deficiency head, a severity head, an image-quality head, and an uncertainty head. A confidence-aware decision system evaluates a prediction reliability, an inter-task consistency, and a contextual validity, and accepts, defers, or withholds the output, escalating a low-confidence prediction to a secondary or a cloud-assisted inference. The shared backbone avoids duplicated computation, reducing a model size, an inference latency, and a battery consumption relative to separate models, while resource-monitoring and model-selection maintain operability across varying hardware. The system operates fully offline, detecting a disease and a deficiency within one inference cycle, in place of a conventional, subjective, or connectivity-dependent diagnostic arrangement.
Disclaimer: Curated by HT Syndication.
