Automated Labeling Device for Machine Learning Data Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The workload for labeling large amounts of learning data is excessively high for labeling persons, even when using tools like 'labelling' for object detection and 'Labelbox' for scene segmentation.
Innovation Solution
A labeling device that includes image-signal acquisition circuitry, machine learning-based image recognition circuitry, and learning-data-set generation circuitry to automate or semi-automate the labeling process by generating learning data sets based on image recognition results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual labeling is performed by labeling persons, then labeling accuracy can be maintained, but the workload and time consumption increase significantly
Solution Approach 1:
The system performs preliminary image recognition and object detection before the labeling process, pre-identifying objects and their boundaries in the images. This preliminary action provides a head start for the labeling process, reducing the time and effort needed for manual labeling while maintaining accuracy.
Solution Approach 2:
The system introduces an intermediary automated labeling process that acts as a bridge between raw images and final labeled data. This intermediary process uses image recognition algorithms to generate initial labels, which then serve as a foundation for further refinement, reducing the overall labeling workload.
2Extent of automation
If automated image recognition is used, then labeling workload is reduced, but system complexity increases
Solution Approach 1:
The system segments the labeling process into distinct functional modules: image acquisition, image recognition, object detection, and label generation. Each module performs a specific task independently, which simplifies the overall system architecture and makes it easier to manage and maintain despite the high level of automation.
Solution Approach 2:
The system employs a universal image recognition framework that can handle multiple types of objects and labeling tasks simultaneously. This multi-functional approach reduces the need for separate specialized systems for different labeling scenarios, thereby managing complexity while maintaining high automation levels.
3Productivity
If manual labeling is performed, then data quality can be ensured, but productivity decreases
Solution Approach 1:
The system incorporates feedback mechanisms where the automated labeling results are evaluated and refined. The image recognition system learns from the labeling outcomes and adjusts its parameters to improve accuracy over time, ensuring that data quality is maintained while increasing productivity through automation.
Solution Approach 2:
The system dynamically adapts its labeling process based on the characteristics of the input images and the performance of the image recognition algorithms. It adjusts parameters such as detection thresholds and confidence levels in real-time to optimize both productivity and labeling accuracy for different datasets.
Data Source
AI summary
A labeling device includes: an image-signal acquisition unit that acquires an image signal indicating an image captured by a camera; an image recognition unit that has learned by machine learning and performs image recognition on the captured image; and a learning-data-set generation unit that generates, by performing labeling on each object included in the captured image on the basis of a result of image recognition, a learning data set including image data corresponding to each object and label data corresponding to each object.