MUMBAI, India, Aug. 31 -- Intellectual Property India has published a patent application (202641101750 A) filed by Jss Science And Technology University on August 23, 2026, for System And Method For Automated Multimodal Ai-Based Chronological 3d Scene Reconstruction And Xr Streaming.
Inventors include Vijay Mallupura Basavaraju; and Mohit Raj.
The application for the patent was published on August 28, 2026, under issue no. 35/2026.
Abstract: ABSTRACT SYSTEM AND METHOD FOR AUTOMATED MULTIMODAL AI-BASED CHRONOLOGICAL 3D SCENE RECONSTRUCTION AND XR STREAMING A system and method for automated, verified, chronological three-dimensional scene reconstruction from unstructured text is disclosed. An ingestion module parses textual records, via a large language model, into a structured interaction timeline. A spatial-grounding module binarizes an environment scan into an orthographic map, transforms two- dimensional coordinates into three-dimensional world coordinates via a defined transformation, and aligns generated assets to the scan through collider-based ray-casting. A cross-modal verification engine projects rendered-scene snapshots and source text into a shared embedding space, computes a similarity score, and invokes a human-in-the-loop correction interface below a threshold of substantially 0.85, reducing hallucination from 31.6% to 2.8% while achieving 97.2% accuracy. A navigation-transformer and quantized code- generation model synthesize executable animation instructions without manual keyframing. A distributed architecture renders on a host device and delivers volumetric output to resource-constrained mobile headsets over a secure, authenticated tunnel, sustaining 62- 72 frames per second. Fig. 1
Disclaimer: Curated by HT Syndication.
