Systems, media, and methods to augment user interactions with visual attention and computer vision models

The system addresses the imprecision of VAS by automatically segmenting and refining AOIs, ensuring accurate visual attention metrics through user-defined prompts, enhancing precision and efficiency in visual content analysis.

WO2026146385A1PCT designated stage Publication Date: 2026-07-093M INNOVATIVE PROPERTIES CO
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
3M INNOVATIVE PROPERTIES CO
Filing Date
2025-12-23
Publication Date
2026-07-09

AI Technical Summary

Technical Problem

Current visual attention software (VAS) systems fail to precisely align areas of interest (AOIs) with the actual entities of interest, leading to inaccurate and diluted attention metrics due to inclusion of surrounding background and adjacent objects.

Method used

A system and method that automatically executes a segmentation model on user-defined AOIs, classifies entities, and allows user refinement through inclusion or exclusion prompts to optimize visual attention goals, using a computing device with processors and memory to enhance precision and efficiency.

Benefits of technology

Generates precise pixel-level segmentation of target entities, reducing user inputs and streamlining workflows, enabling accurate visual attention metrics and context-specific modifications for scene optimization.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure IB2025063408_09072026_PF_FP_ABST
    Figure IB2025063408_09072026_PF_FP_ABST
Patent Text Reader

Abstract

A computer‑implemented method includes receiving image data and one or more user-defined areas of interest (AOIs). The method further includes automatically executing a segmentation model on each AOI in response to a trigger and classifying each segmented entity to assign a class type including at least text or object. The method additionally includes receiving user input that includes inclusion point prompts, exclusion point prompts, masks of multiple points, or any combination thereof, to refine the segmentation. The method also further includes optimizing a scene, based at least in part on the user input, to achieve visual attention goals.
Need to check novelty before this filing date? Find Prior Art