2D to 3D Feature Conversion for Low-Cost Training Data

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The high cost associated with creating training data for systems that detect human behavior, as existing methods often require specialized equipment like motion capture systems.

Innovation Solution

An image processing device and method that extracts two-dimensional features from images, converts them into three-dimensional features, and generates training data using these features along with labels indicating a person's physical state, without the need for special equipment like motion capture systems.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If motion capture systems or special equipment are used to create training data, then the accuracy and quality of training data are improved, but the cost increases

Engineering Contradiction:

Improvetraining data qualityVSAvoidcost

Core Design Contradiction:

Measurement precisionVSEase of manufacture

Solution Approach 1:

The patent uses 2D images captured by ordinary cameras as copies or projections of the actual 3D scene, and through image processing algorithms reconstructs 3D information. This copying approach avoids the need for expensive motion capture equipment while still generating accurate training data for machine learning models.

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent replaces mechanical motion capture systems with computational image processing methods. Instead of using specialized mechanical equipment to capture 3D motion data, the system uses 2D images processed through algorithms to extract 3D features, substituting a computational approach for a mechanical one.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Ease of manufacture

If 2D images are processed to generate training data without special equipment, then the cost is reduced, but the accuracy of three-dimensional feature extraction may deteriorate

Engineering Contradiction:

ImprovecostVSAvoidthree-dimensional feature accuracy

Core Design Contradiction:

Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent transforms 2D image data into 3D feature representations through computational processing. By adding a dimensional transformation step, the system recovers depth and spatial information that is lost in 2D images, enabling accurate 3D feature extraction without requiring 3D capture equipment.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The patent introduces image processing algorithms as an intermediary between 2D images and 3D feature extraction. These algorithms act as a mediator that bridges the gap between 2D image data and the required 3D information, enabling accurate feature extraction through computational methods rather than direct 3D measurement.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS11803615B2Generating 3D training data from 2D images

Publication Date: 2023.10.31 NEC CORP

AI summary

An image processing device includes an extraction unit configured to extract a two-dimensional feature regarding a part of a person in an image, a conversion unit configured to convert the two-dimensional feature into a three-dimensional feature regarding a human body structure, and a training data generation unit configured to generate training data using the three-dimensional feature and a label indicating a physical state of the person.