Clustering-Based Labeling for Fast Product Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In the smart factory field, manual labeling of products for anomaly detection and classification of good/defective products is burdensome due to the high volume and speed of production, requiring significant time and cost for data labeling to build machine learning models.
Innovation Solution
A device and method utilizing clustering techniques to automatically or semi-automatically perform labeling by mapping input data into a vector space, creating clusters, and performing connected component analysis to assign labels based on edge values, allowing for anomaly detection without extensive manual labeling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual labeling is performed for all products to build machine learning models, then labeling accuracy is improved, but labeling time and cost increase significantly
Solution Approach 1:
The patent segments the labeling process into two parts: (1) automatic clustering-based labeling for the majority of products, and (2) manual labeling only for ambiguous cases that fall into multiple clusters. This segmentation allows most products to be labeled quickly automatically while maintaining accuracy by having experts review only the uncertain cases.
Solution Approach 2:
The system performs self-labeling by automatically clustering product data and assigning labels based on cluster characteristics. The algorithm autonomously identifies patterns and assigns labels without human intervention for clear cases, reducing the burden on manual labelers to only handle ambiguous situations.
2Measurement precision
If manual labeling is performed for all products to build machine learning models, then labeling accuracy is improved, but production speed decreases
Solution Approach 1:
The labeling process is segmented into automatic clustering for high-speed processing and manual review only for ambiguous cases. This allows the system to maintain high production speed for the majority of products while ensuring accuracy through targeted manual verification of uncertain cases.
Solution Approach 2:
The system changes the parameter of labeling speed by introducing automatic clustering algorithms that can process products at high speed. The clustering parameters (number of clusters, distance thresholds) are optimized to balance speed and accuracy, allowing rapid processing while maintaining sufficient labeling quality.
3Measurement precision
If extensive manual labeling is performed to build machine learning models, then model accuracy is improved, but cost increases significantly
Solution Approach 1:
The patent segments the labeling workload by cost, assigning automatic low-cost clustering to the majority of products and reserving expensive manual expert labeling only for ambiguous cases that require human judgment. This segmentation dramatically reduces overall labeling costs while maintaining model accuracy through targeted expert review.
Solution Approach 2:
Instead of performing full manual labeling on all products, the system applies partial manual action only where necessary (ambiguous cases). This partial action approach reduces costs significantly while still achieving sufficient model accuracy by focusing expert resources on the most challenging cases.
4Productivity
If automatic clustering-based labeling is used, then labeling time and cost are reduced, but labeling accuracy may decrease
Solution Approach 1:
The system incorporates feedback mechanisms where the results of automatic clustering are evaluated, and ambiguous cases are identified and fed back for manual review. This feedback loop allows the system to correct errors and improve accuracy iteratively, ensuring that automatic labeling does not compromise overall labeling quality.
Solution Approach 2:
The patent introduces an intermediary manual review process that acts as a mediator between automatic clustering and final labeling. The intermediary reviews ambiguous cases and makes final decisions, ensuring that automatic labeling efficiency does not compromise accuracy by catching and correcting potential errors.
Data Source
AI summary
A method for labeling comprises: a clustering unit forms a plurality of nodes, upon the input of a plurality of pieces of data, by projecting the plurality of pieces of input data onto a predetermined vector space; the clustering unit generates clusters by clustering the plurality of nodes; a labeling unit carries out connected component analysis of the generated clusters to derive one or more connected components; and the labeling unit labels the connected components. In addition, an anomaly detection method based on clustering of the present invention comprises the step in which a detecting unit determines whether difference between a mock data cluster and an input data cluster is equal to or greater than a preset threshold value and, if the difference is determined to be equal to or greater than a preset threshold value, determines that there is an anomaly in the input data cluster.