Photo-Realistic Image–Model Mapping for Accurate 3D Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated classification techniques for photo-realistic images are imprecise and time-consuming, particularly when large collections of images need to be tagged, and there is a need to improve the accuracy of these classifications.
Innovation Solution
The technique adjusts the classification of a photo-realistic image by leveraging the classifications from other images that correspond to the same portion of a 3D model generated from the photo-realistic images, using methods such as Bayes' formula and weight adjustments based on spatial metadata, to enhance the accuracy of classifications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automated classification techniques are used to tag photo-realistic images, then productivity is improved, but measurement precision deteriorates
Solution Approach 1:
The patent combines multiple image classifications from different images that depict the same physical object or scene into a single aggregated classification. By merging classification results across multiple images, the system achieves both automated processing efficiency and improved classification accuracy through collective intelligence of multiple views.
Solution Approach 2:
The system uses feedback from multiple image classifications to refine and adjust the final classification result. Each image classification provides feedback information that contributes to the aggregated result, allowing the system to iteratively improve accuracy by considering evidence from multiple sources before finalizing the classification.
2Measurement precision
If manual tagging is used to classify photo-realistic images, then measurement precision is improved, but productivity deteriorates
Solution Approach 1:
The system performs self-service by automatically aggregating and refining classifications without requiring manual intervention. The automated system serves itself by taking multiple automated classifications, combining them through aggregation algorithms, and producing improved results, thereby maintaining high productivity while enhancing accuracy.
3Productivity
If machine learning engines are used to automatically identify objects, then productivity is improved, but measurement precision deteriorates due to indefinite classifications
Solution Approach 1:
The patent merges multiple machine learning classifications that may each have low individual confidence into a single aggregated classification with higher overall confidence. By combining evidence from multiple independent machine learning engines processing different images, the system achieves both automation and improved precision.
Solution Approach 2:
The system changes the parameter of classification confidence by aggregating results across multiple images. The aggregation process transforms individual low-confidence classifications into a high-confidence collective classification, effectively changing the confidence parameter through mathematical combination of multiple measurements.
Data Source
AI summary
Techniques are provided for increasing the accuracy of automated classifications produced by a machine learning engine. Specifically, the classification produced by a machine learning engine for one photo-realistic image is adjusted based on the classifications produced by the machine learning engine for other photo-realistic images that correspond to the same portion of a 3D model that has been generated based on the photo-realistic images. Techniques are also provided for using the classifications of the photo-realistic images that were used to create a 3D model to automatically classify portions of the 3D model. The classifications assigned to the various portions of the 3D model in this manner may also be used as a factor for automatically segmenting the 3D model.