Systems and methods for generating dimensionally coherent training data

By transforming 2D images into 3D visualizations and perturbing 3D representations, the method generates dimensionally coherent training data that addresses inaccuracies in feature matching, improving network performance and reducing computational costs.

AU2023225698B2Pending Publication Date: 2026-07-23HOVER INC
View PDF 3 Cites 0 Cited by

Patent Information

Authority / Receiving Office
AU · AU
Patent Type
Applications
Current Assignee / Owner
HOVER INC
Filing Date
2023-02-23
Publication Date
2026-07-23

AI Technical Summary

Technical Problem

Existing 3-D reconstruction methods generate training data that is not dimensionally coherent, leading to inaccurate feature matching and increased computational costs due to the use of 2D data formats that fail to preserve 3D relationships and introduce variability, resulting in false positives and negatives when deployed in real-world scenarios.

Method used

Transform 2D images into 3D visualizations that preserve spatial relationships and generate synthetic views by perturbing 3D representations, using visibility masks and photo-texturing to create training datasets that maintain accurate 3D geometries and relationships.

Benefits of technology

Improves the accuracy of feature matching by training networks on spatially coherent data, reducing computational costs and enhancing the network's ability to recognize features across different camera views.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 00000001_0000
    Figure 00000001_0000
  • Figure 00000029_0000
    Figure 00000029_0000
  • Figure 00000029_0001
    Figure 00000029_0001
Patent Text Reader

Abstract

System and method are provided for generating training data for feature matching among images of a building structure. The method includes obtaining a model of a building that includes a camera solution and images used to generate the geometric model. The method also includes, for facades of the model: applying a minimum bounding box to a respective facade to obtain a respective facade slice that is a 2-D plane represented in a 3-D coordinate system of the model; and projecting visual data of at least one camera in the camera solution that viewed the respective facade onto a visibility mask associated with the respective facade slice. The method also includes photo-texturing the projected visual data facade slice to one of the facade slices or the geometric model to generate a visual 3-D representation of the building; and generating a training dataset by perturbing the visual 3-D representation.
Need to check novelty before this filing date? Find Prior Art