MUMBAI, India, Oct. 26 -- Intellectual Property India has published a patent application (202621106800 A) filed by Vishwakarma University on September 04, 2026, for An Adaptive Deep Learning-Based Intelligent Pattern Recognition And Classification System And Method Thereof.
Inventors include Dr. Anjali Shailesh Sardesai; Prof. Anjali Landge; Prof. Khushbu Kadukar; Dr. Bharati Pandurang Kawade; Prof Abdul Samad Riyaz Ahmed Khan; Dr. Ganesh Chandrakant Gadekar; Dr. Swapnil Purushottamrao Goje; Prof. Santosh Bhauso Takale; and Dr. Anjali Motwani.
The application for the patent was published on October 23, 2026, under issue no. 43/2026.
Abstract: An adaptive deep learning-based intelligent pattern recognition and classification system and method are disclosed. The system includes an input acquisition interface, a multi-resolution encoder, a feature-quality estimator, an adaptive feature-routing network, a prototype-memory classification engine, an uncertainty calibration and rejection engine, a context and distribution-drift monitor, and a resource-aware inference controller. Multi-resolution latent representations are evaluated for quality and selectively routed so that unreliable feature subspaces are suppressed and computing effort is allocated according to input difficulty and available resources. Classification combines learned class evidence with similarity to bounded adaptive class prototypes. Prototype updates are gated by confidence, consistency, distance, and drift conditions. Calibrated uncertainty prevents forced assignment of unfamiliar or corrupted patterns to known classes. Detected distribution drift modifies routing, confidence, adaptation, or inference depth. The disclosed architecture thereby improves robustness, computational efficiency, and controlled handling of unknown patterns in deployed pattern-recognition systems. Accompanied Drawing [FIGS. 1-2]
Disclaimer: Curated by HT Syndication.
