Active Learning Product Inspection Engine
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional visual inspection methods in manufacturing facilities, relying on human vision, are limited and prone to errors due to intra-inspector and inter-inspector variability, and require significant human effort for training data annotation, especially for minor changes or new product designs.
Innovation Solution
A machine learning-trained product inspection engine is actively trained using automated identification and labeling of new training images based on inspection results, with human annotation data augmentation, enabling efficient generation of training data and improved model accuracy over time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If human vision is used for visual inspection, then flexibility and adaptability to new products are maintained, but inspection accuracy deteriorates due to intra-inspector and inter-inspector variability
Solution Approach 1:
The patent replaces the mechanical human visual inspection system with an automated machine learning-based inspection engine. The system uses trained models to analyze product images, eliminating human variability while maintaining adaptability through continuous model training with new product data. This substitution resolves the contradiction by providing both high precision (through automated consistent analysis) and adaptability (through retrainable models).
Solution Approach 2:
The inspection system implements self-service through automated model training and improvement. The system automatically trains its own machine learning models using inspection data, continuously improving its accuracy without requiring manual reprogramming for each new product. This self-training capability enables the system to maintain both high inspection accuracy and adaptability to new products simultaneously.
2Reliability
If manual annotation of training data is performed, then model training can be customized for specific products, but time consumption and labor effort increase significantly
Solution Approach 1:
The system performs self-service by automatically generating training data annotations through its own inspection processes. Inspection images and their corresponding results are automatically used to create training datasets, eliminating the need for manual annotation. This self-annotation capability maintains high model training accuracy while dramatically reducing the time and labor required for data preparation.
Solution Approach 2:
The system implements feedback loops where inspection results are automatically fed back into the training data generation process. The inspection engine analyzes products, and these same images with automated labels are used to retrain and improve the models. This feedback mechanism enables continuous model improvement using automatically generated training data, resolving the contradiction between training reliability and time consumption.
3Productivity
If automated inspection is implemented, then inspection speed and consistency are improved, but the system requires extensive initial training data and setup time
Solution Approach 1:
The system implements dynamics by making the training data requirement flexible and adaptive. Rather than requiring fixed extensive initial datasets, the system can dynamically adjust its training needs based on available data and continuously improve through ongoing operation. The model training process is dynamic, allowing the system to start with limited data and progressively improve as more inspection data becomes available, reducing initial setup complexity while maintaining high inspection speed.
Solution Approach 2:
The system performs preliminary actions by using automatically generated training data from initial inspections to prepare models before full-scale deployment. Rather than requiring extensive pre-collected training data, the system uses its own initial inspection results to create training datasets, enabling rapid deployment with minimal initial setup while maintaining high inspection productivity.
Data Source
AI summary
A computing entity is described that obtains at least one inspection image of an at least partially fabricated product and causes the at least one inspection image to be processed by a product inspection engine. The product inspection engine includes a machine learning-trained model. The computing entity obtains an inspection result determined based on the processing of the at least one inspection image by the product inspection engine; identifies one or more training images stored in an image database based at least in part on the at least one inspection image; associates automatically generated labeling data with the one or more training images based at least in part on the inspection result determined by the processing of the at least one inspection image; and causes training of the product inspection engine using the one or more training images and the associated labeling data.