RPA Data Labelling from Production Sensor Workflows
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The high cost and labor intensity of generating labelled datasets for machine learning models, particularly in industrial production settings, where traditional data labelling methods are time-consuming and prone to human errors, and existing approaches do not efficiently address the challenges of data collection and labelling quality.
Innovation Solution
Implementing a system that uses sensors to capture data during normal production processes, leveraging expert workers' classifications to create labelled datasets, and deploying machine learning models to automate classification tasks, while continuously improving model accuracy through additional training and data collection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional manual data labelling methods are used, then data quality can be maintained through human expertise, but the process becomes expensive and labor intensive
Solution Approach 1:
The system allows workers to perform their normal duties while sensors automatically capture and label their actions, eliminating the need for dedicated data collection tasks. The production process itself generates the labelled data as a byproduct, making the system self-servicing for data collection purposes
Solution Approach 2:
The sensor system serves multiple functions: it monitors production quality, tracks worker performance, and simultaneously generates labelled training data for machine learning models. This multi-functionality resolves the contradiction by making the data collection process efficient without compromising quality
2Reliability
If more labelled data is collected to improve model accuracy, then model performance improves, but the cost and time for data collection increases
Solution Approach 1:
Sensors continuously capture data during normal production operations without interruption. The system operates continuously as workers perform their duties, accumulating labelled data over time without dedicated data collection periods, thus improving model accuracy without proportional time loss
Solution Approach 2:
The system pre-captures labelled data during production processes before it is needed for model training. By having data already labelled and stored in advance, the system eliminates the time delay that would occur if data were collected after model development begins
3Measurement precision
If manual data labelling is performed to ensure data quality, then labelling accuracy improves, but the process becomes prone to human errors and is time-consuming
Solution Approach 1:
The system replaces manual human labelling with automated sensor-based detection and classification. Sensors objectively capture and label worker actions without human intervention in the labelling process itself, eliminating human errors while maintaining accuracy, and reducing process complexity by automating what was previously a manual task
4Quantity of substance
If traditional data collection methods are used, then data can be obtained for training, but the process is expensive and labor intensive
Solution Approach 1:
The production system generates labelled training data as a self-service byproduct of normal operations. Workers do not need to be trained as data annotators, and no separate data collection team is required. The system automatically produces the training data needed, increasing efficiency while maintaining data quantity
Solution Approach 2:
The system merges the data collection function with the existing production process. Instead of separating data collection from production, the system combines them so that production activities simultaneously generate the labelled data needed for training, thereby increasing productivity without sacrificing data quantity
Data Source
AI summary
One application of deep learning methods and labelled data is for industrial production or work applications. For such applications implemented with machine learning applications, massive amounts of data are required to train, validate, and/or tune models for better fitting the requirements. However, obtaining such data has typically be costly and difficult. Embodiments provide adaptable processes that provide data labelling methods for work settings. Embodiments take advantage of the work or production processes to label and collect data, which save time and money and improves accuracy. Embodiments prevent or reduce the need for worker training costs and human mistake-triggered data labelling problems. Embodiments also improve data labelling quality and speed-up of the development cycle.