MUMBAI, India, Aug. 10 -- Intellectual Property India has published a patent application (202641090011 A) filed by Retearn Technologies Private Limited on July 24, 2026, for Method And System For Ai-Assisted Annotation Using Exclusion-Guided Segmentation And Rule-Based Classification.
Inventors include Tirumala Vangipuram Sujan Parthasaradhi; Vivek; Waghmare Sanket Laxman; Pasham Nithish Reddy; and Kandukuri Sai Swaroop Reddy.
The application for the patent was published on July 31, 2026, under issue no. 31/2026.
Abstract: ABSTRACT Method and System for AI-Assisted Annotation Using Exclusion-Guided Segmentation and Rule-Based Classification The present disclosure relates to a novel method and system for multi-layered AI-powered annotation of composite objects by way of exclusion-guided segmentation and rule-based & AI-powered classification. In this context, user inputs are taken to produce segmentations from an AI/ML model, where the annotations that have been validated are used as constraints and negative prompts in the exclusion-based segmentation process so that no redundant segmentation takes place. According to the invention, a novel process allows for decomposition of composite objects into primary & sub-components in sequential order without overlap. In the classification layer, the semantic classes assigned to segments by means of rules and AI are accompanied by application-specific processing rules, e.g., merging primary object segments and retaining sub-component segments individually. Additionally, the invention creates annotation results as polygons, normalized bounding boxes, and resolution-independent annotations with metadata. Atomic storage mechanisms are provided to ensure proper validation and storage of annotation data.
Disclaimer: Curated by HT Syndication.
