Composite Image Training for Accurate Object-User Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies struggle to accurately identify which object is affected by a person's motion in scenes with numerous similar objects or complex backgrounds, leading to decreased detection accuracy.
Innovation Solution
A generation program and method that generates composite image data by arranging extracted objects near a person in a specific position, using contrastive learning to train a machine learning model to enhance object identification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional detection rules or basic machine learning models are used to detect object interactions, then the system can operate with simple processing, but the detection accuracy decreases in scenes with numerous similar objects or complex backgrounds
Solution Approach 1:
The system performs preliminary actions by generating composite images that arrange multiple objects in various configurations before training the machine learning model. This pre-processing step creates diverse training data that helps the model learn to distinguish between similar objects in complex scenes, thereby improving detection accuracy without requiring complex processing during actual detection operations
Solution Approach 2:
The system dynamically adjusts the training process by varying object arrangements, positions, and configurations in composite images. This dynamic approach allows the model to adapt to different scene complexities and object similarities, improving its ability to accurately detect object interactions in diverse environments while maintaining efficient processing
2Measurement precision
If detailed detection rules are set in advance to improve detection accuracy, then the accuracy improves for specific scenarios, but the system becomes less adaptable to different arrangements and orientations
Solution Approach 1:
The system creates a universal machine learning model through composite image training that can handle multiple object arrangements, orientations, and scene configurations. The model learns general patterns of object interactions rather than scenario-specific rules, enabling it to adapt to various detection scenarios including different camera angles, person orientations, and object placements without requiring reconfiguration
Solution Approach 2:
The system varies multiple parameters in composite image generation including object positions, distances, orientations, and scene backgrounds. This parameter variation during training enables the model to learn robust detection capabilities across different conditions, maintaining both high accuracy and adaptability to new scenarios
3Measurement precision
If machine learning models are trained with basic image data, then the training process is simple, but the model cannot accurately identify which object is affected by person's motion in complex scenes
Solution Approach 1:
The system performs preliminary action by generating composite images with multiple objects arranged in various configurations before model training. This pre-processing creates rich training data that teaches the model to distinguish between similar objects and accurately identify which object is being interacted with, improving object identification accuracy in complex scenes
Solution Approach 2:
The system creates copies of objects in composite images to simulate complex scenes with multiple similar objects. By training on these copied and arranged objects, the model learns to differentiate between them and accurately identify the target object of interaction, enhancing its performance in scenes with numerous similar items
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An information processing apparatus extracts an object that is used by a person included in an image by acquiring the image that includes the person and analyzing the acquired image. The information processing apparatus generates a composite image in which the extracted object is arranged at a position that satisfies a predetermined condition on a basis of a position of the object that is used by the person included in the acquired image. The information processing apparatus generates, by using the generated composite image, a machine learning model that has been trained to identify the person who uses the object.