Sensor Event Detection Retraining With Synthetic Training Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The high cost of collecting training data for generating inference models is a significant challenge in existing sensor-based event detection systems, particularly due to the need for manual data collection and labeling across varying environments.
Innovation Solution
An information processing system that automatically generates training data using simulation techniques based on sensor metadata and environmental conditions, allowing for retraining of inference models to improve detection accuracy without manual intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual data collection and labeling is performed across varying environments, then detection accuracy can be improved, but the cost and time required increase significantly
Solution Approach 1:
The patent creates synthetic copies of real-world sensor data through simulation. The simulation apparatus generates artificial sensor data and labeled detection information by virtualizing real-world scenarios, thereby obtaining training data without the need for manual collection and labeling of actual sensor data across varying environments.
Solution Approach 2:
The patent replaces the mechanical process of manual data collection and labeling with an automated simulation system. Instead of physically capturing images and manually annotating them, the system uses computational simulation to generate training data automatically, substituting human labor with an automated virtual environment.
2Measurement precision
If manual data collection and labeling is performed across varying environments, then detection accuracy can be improved, but the cost increases significantly
Solution Approach 1:
The patent creates synthetic copies of real-world sensor data through simulation. The simulation apparatus generates artificial sensor data and labeled detection information by virtualizing real-world scenarios, thereby obtaining training data without the need for manual collection and labeling of actual sensor data across varying environments.
Solution Approach 2:
The patent replaces the mechanical process of manual data collection and labeling with an automated simulation system. Instead of physically capturing images and manually annotating them, the system uses computational simulation to generate training data automatically, substituting human labor with an automated virtual environment.
3Measurement precision
If inference models are trained for specific environments, then detection performance in those environments improves, but the system loses adaptability to new environments
Solution Approach 1:
The patent implements dynamic adaptability by continuously generating and using simulation data that represents varying environmental conditions. The simulation apparatus can generate training data for different environments on demand, allowing the system to adapt to new environments without requiring re-collection of real-world data, thus maintaining both specialized performance and general adaptability.
Data Source
AI summary
It is desirable that a technique that allows a reduction in cost of collecting training data required for generating an inference model be provided.Provided is an information processing device including: a training data generation unit configured to generate second training data on the basis of the fact that detection information related to detection of a predetermined event does not satisfy a first condition, the detection information being obtained on the basis of a first inference model generated through training based on first training data and sensor data detected by a sensor; and a retraining unit configured to perform retraining on the basis of the second training data to obtain a second inference model.