Vehicle Sensor Data Collection With Real-Time Labeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for collecting data for machine learning, such as those involving vehicles and environmental sensors, do not allow for immediate interaction with raw data, leading to inefficient and time-consuming data preparation, which affects the speed and accuracy of machine learning processes, particularly when using neural networks trained on big data.
Innovation Solution
A method and system for collecting data using environmental sensors that interact with raw data in real-time, allowing for immediate marking up, modifying, deleting, or not writing data to a machine-readable storage medium, thereby speeding up the preparation of data suitable for machine learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If raw data are collected and stored without immediate interaction, then data volume is maximized, but data preparation time increases significantly
Solution Approach 1:
The system performs preliminary actions by immediately marking up, modifying, filtering, or deleting raw data at the point of collection before storage. This preliminary processing reduces the burden of data preparation later, allowing the system to maintain large data volumes while significantly reducing the time needed for subsequent data preparation and model training.
2Reliability
If all raw data are stored without filtering or processing, then data completeness is maintained, but machine learning efficiency decreases
Solution Approach 1:
The system applies different processing operations to different portions of raw data based on local quality requirements. Some data are marked up with annotations, some are filtered to remove noise, some are deleted if irrelevant, and some are stored in original form. This selective processing maintains the reliability and completeness of necessary data while improving machine learning efficiency by reducing the overall data preparation burden.
3Ease of manufacture
If data processing is delayed until after collection, then data collection simplicity is maintained, but overall system speed decreases
Solution Approach 1:
The system merges the data collection process with immediate data processing operations. The data collection device performs marking up, modifying, filtering, and deleting operations in real-time during data collection, rather than as separate subsequent steps. This integration maintains the simplicity of data collection while significantly accelerating the overall system speed by eliminating delays between collection and processing.
Data Source
AI summary
The technical solution relates to the field of transport, in particular, to auxiliary devices and methods for collecting data for machine learning using vehicles, such as, for example, personal mobility devices (PMD). There is a need to speed up the preparation of data suitable for machine learning and thus reduce the time it takes to prepare machine learning models trained on big data. The technical result achieved by implementing the claimed technical solution, in addition to the implementation of the product and/or method for its intended purpose, is an increase in the speed of preparing data suitable for machine learning, including an increase in the speed of preparing big data that can be used for machine learning. In some aspects, another technical result achieved is also an increase in road safety.