Automated Labeling for Autonomous Driving Training Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The generation of training data for automated driving requires a significant amount of effort, cost, and time due to the extensive annotation process needed for machine learning.
Innovation Solution
A data generation apparatus that automates the labeling of external environment information, using sensors to collect data and a server to apply labeling processing, reducing the manual effort required for annotation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual annotation is used for training data generation, then labeling accuracy can be ensured, but the effort, time, and cost required become enormous
Solution Approach 1:
The system uses the vehicle's own travel data and sensor information to automatically generate labels for training data. The labeling is performed based on the vehicle's actual travel trajectory and environmental sensor readings, eliminating the need for manual annotation while maintaining accuracy through self-generated ground truth data
Solution Approach 2:
The manual mechanical process of annotating training data is replaced with an automated computer-based system that uses sensor fusion and trajectory analysis to generate labels automatically, substituting human labor with algorithmic processing
2Reliability
If manual annotation is used for training data generation, then label quality can be controlled, but the cost and effort required become enormous
Solution Approach 1:
The system generates its own training data labels automatically using its sensor suite and travel information, making the data generation process self-sufficient and eliminating external annotation services, thereby reducing cost while maintaining quality through consistent automated processing
Solution Approach 2:
The vehicle's sensor system serves multiple functions: it collects data for environmental perception during normal operation and simultaneously generates labeled training data for machine learning, eliminating the need for separate data collection and annotation processes
3Measurement precision
If extensive annotation is performed for machine learning, then model training accuracy improves, but the productivity of data generation decreases
Solution Approach 1:
The system continuously generates labeled training data during normal vehicle operation without interruption. The annotation process occurs continuously as the vehicle travels, converting raw sensor data into labeled training examples in real-time, thereby maintaining both data quality and generation speed
Solution Approach 2:
The system performs data collection and preliminary labeling during the vehicle's normal travel before the actual machine learning training process. This preliminary preparation of labeled data ensures that high-quality training examples are ready in advance, improving both accuracy and overall process efficiency
Data Source
AI summary
A data generation apparatus for automated travel, the data generation apparatus being a data collection apparatus characterized by comprising: obtaining means for obtaining external environment information; and labeling means for adding, to focus information included in the external environment information obtained by the obtaining means, a label corresponding to passing of a vehicle through a position at which the external environment information has been collected.