MUMBAI, India, July 7 -- Intellectual Property India has published a patent application (202621055328 A) filed by Ojassingh Yogeshsingh Bayas on April 30, 2026, for Ar Guided Multi-Modal Belief-Driven Diagnostic System Using Visual Embeddings And Sentence-Transformer Semantics With Fine-Tuned Yolov8.
Inventors include Sakshi Ganesh Doifode; Anshaman Satish Athaley; Pavan Sanjay Bhalerao; Ojassingh Yogeshsingh Bayas; Aryan Amit Sonawane; and Premanand Ghadekar.
The application for the patent was published on July 03, 2026, under issue no. 27/2026.
Abstract: The present invention discloses an AR-guided multi-modal belief-driven diagnostic system that combines context-conditioned BLIP-2 visual symptom extraction, Sentence-BERT semantic embeddings, Bayesian belief propagation, and information-theoretic question selection to automate hardware and software troubleshooting for computing devices. The system fuses expert-curated rules with dynamically learned symptom-cause patterns via a confidence-weighted belief engine, selects diagnostic questions that maximise expected information gain to minimise user interaction, and retrieves ranked repair tutorials through a hybrid dense-sparse vector search with feedback re-ranking. A self-learning engine mines resolved sessions nightly using FP-Growth pattern discovery to autonomously expand the knowledge base. Fine-tuned category-specific YOLOv8 models provide hardware component detection. Results demonstrate 91.2% diagnostic accuracy, an average of 3.7 questions per session (40.3% reduction over baseline), a 73% reduction in time-to-diagnosis, and autonomous knowledge base growth of 147% over 90 days.
Disclaimer: Curated by HT Syndication.
