MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202641094118 A) filed by M. V. Karthikeya; Dr. T. V. Nagalakshmi; Somineni Sri Varshini; Pranshu Parth; and Durga Aditya Shankar Prasad Pandey on August 04, 2026, for System And Method For Secure, Session-Gated, Real-Time Indoor Object Detection Employing Route- Independent Authentication Gating And Domain-Curated Semantic Class-Whitelist Filtering.
Inventors include M. V. Karthikeya; Dr. T. V. Nagalakshmi; Somineni Sri Varshini; Pranshu Parth; and Durga Aditya Shankar Prasad Pandey.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: A system and method are disclosed for secure, real-time indoor object detection and authenticated video streaming. Successive video frames captured from an image- capture device are processed by a trained convolutional neural network to yield candidate object detections. A semantic class-whitelist filtering unit compares each candidate detection's classified category against a predetermined, domain-curated whitelist of indoor-appropriate categories, discarding detections outside that whitelist so as to reduce context-inappropriate misclassification without retraining the underlying network. Retained detections are rendered onto the frame, encoded, and assembled into a continuously updating multipart video stream. A session-validation gate, coupled to an authentication module issuing cryptographically signed session credentials, is invoked independently upon every network-reachable route capable of exposing video content or inference output, including the continuous stream itself and a discrete-image-analysis route, and completes that verification before any worker thread or inference resource is allocated, such that no frame, detection result, worker thread, or inference capability is committed to, or disclosed to, a party lacking a valid session credential, and such that the streaming route cannot be reached by bypassing an initial login page. A per-connection, threaded serving architecture permits concurrent servicing of multiple authenticated viewers from a shared, singly- loaded neural-network instance, without blocking the responsiveness of the serving process.
Disclaimer: Curated by HT Syndication.