MUMBAI, India, Sept. 14 -- Intellectual Property India has published a patent application (202641108232 A) filed by Easwari Engineering College; and Srm Institute Of Science And Technology, Ramapuram Camous on September 09, 2026, for Ai-Based Image Hallucination Detection And Verification Of Vision-Language Models.
Inventors include Tanush S; Sriram M; and Lakshmi Priya V.
The application for the patent was published on September 11, 2026, under issue no. 37/2026.
Abstract: The present invention relates to an artificial intelligence-based image verification system for detecting hallucinated content generated by Vision-Language Models. Vision-Language Models often produce descriptions containing objects, attributes, or relations not actually present in the input image. The system receives an input image 5 and performs preprocessing including image loading, resizing, and normalization. The preprocessed image is independently analyzed using a YOLOv8-based detection module to obtain object labels, bounding boxes, and confidence scores as visual evidence. A Vision-Language Model generates a semantic caption describing the image, which is compared against the independently obtained visual evidence to 10 verify object consistency, unsupported objects, missing objects, counts, attributes, relations, and confidence. A consistency-scoring mechanism determines the degree of agreement, classifying information as verified or potentially hallucinated. Inconsistent captions may subsequently be refined using verified evidence while preserving accurate content. The invention thereby provides a modular, explainable 15 approach for improving the reliability of AI-generated image descriptions. Total Number of Words in Abstract: 150
Disclaimer: Curated by HT Syndication.
