Automated Labelled Dataset Generation Using Proximity Sensors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning models in the transportation industry rely on manual annotation by human operators, which is cumbersome, prone to errors due to distractions or tiredness, leading to inconsistent service quality and potential catastrophic consequences.
Innovation Solution
A system that uses proximity sensors to automatically generate labelled datasets from monitoring sensors, eliminating the need for human annotation, and trains machine learning models to identify and track objects of interest, such as vehicles, for improved decision-making and collision prevention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual annotation by human operators is used, then machine learning models can be trained, but the process is cumbersome and prone to errors due to distractions or tiredness
Solution Approach 1:
The system uses proximity sensors to automatically detect and label objects in monitoring sensor data without requiring human operators. The proximity sensors self-service the labelling task by detecting objects and generating annotations automatically, eliminating human error and fatigue while maintaining high labelling quality and efficiency
Solution Approach 2:
The patent replaces the mechanical human annotation process with an automated sensor-based system. Proximity sensors detect objects and generate labels automatically, substituting human cognitive and manual operations with electronic detection and processing mechanisms
2Reliability
If manual annotation is performed, then training data can be generated, but service quality becomes uncontrollable and non-uniform
Solution Approach 1:
The proximity sensor system performs self-service annotation consistently without human intervention, ensuring uniform service quality. The automated detection and labelling process eliminates the variability inherent in human performance, providing consistent and reliable training data generation
Solution Approach 2:
The system changes the parameters of the annotation process from human-dependent to sensor-dependent. By using proximity sensors with fixed detection criteria and automated processing, the system achieves consistent and reproducible labelling results that maintain uniform service quality
3Measurement precision
If human operators perform labelling, then objects can be identified, but errors occur due to distractions or tiredness
Solution Approach 1:
The proximity sensors continuously and autonomously detect objects without requiring operator attention. The system self-services the object detection and labelling function, maintaining high identification accuracy while eliminating time loss due to operator fatigue and distractions
Solution Approach 2:
The automated sensor system provides continuous object detection and labelling without interruption. Unlike human operators who experience fatigue and distraction, the proximity sensors maintain constant readiness and continuous operation, ensuring uninterrupted and accurate object identification
Data Source
AI summary
A system for generating a labelled dataset is provided. The system comprises processor configured to: receive first data wherein the first data comprises one or more frames and wherein the first data comprises data defining an object of interest within a predetermined area; receive second data wherein the second data is associated with the object of interest within the predetermined area; analyse the one or more frames of the first data to identify, based on the second data, the object of interest present in the first data; label the one or more frames of the first data based on the analysis to generate a labelled dataset; and output the labelled dataset. Also provided is a method for generating a labelled dataset, a system for training a machine learning model, and a detection system for detecting one or more objects of interest.