Method and system for defect inspection based on deep learning
AssigneeHyundai Mobis Co., Ltd.
InventorsTae Hyun KIM, Hye Rin KIM, Yeong Jun CHO
Life of this patent
Estimating the term…
Abstract
The present disclosure in some embodiments provides a product inspection method and a system based on deep learning for detecting a product defect. The present disclosure provides a product inspection method and system for detecting product defects by linking a predeveloped deep learning-based classification model to interwork with the existing product inspection system while fine-tuning the classification model to be maintained or supplemented by instantly correcting errors of the classification model, thereby improving the accuracy of product quality inspection.
Claims11
A method performed by a system for retraining a classification deep learning (DL) model trained to inspect products for defects, comprising:
- receiving a first image capturing a first product;
- providing the received image to an edge computer operating a classification DL model pre-trained to determine, based on a set of parameters, whether or not a product has an acceptable quality based on an image capturing the product, wherein the classification DL model is configured to perform:
- in response to the received first image provided thereto, attempting to determine, based on the first image, whether or not the first product has the acceptable quality;
- determining that whether or not the first product has the acceptable quality cannot be determined using the first image;
- and
- in response to determining that whether or not the first product has the acceptable quality cannot be determined using the first image, determining that the first image contains an error image;
- in response to determining that the first image contains an error image, adjusting, based on the error image, the set of parameters of the classification DL model;
- and
- extracting, from the edge computer, the adjusted set of parameters of the classification DL model.
A system for retraining a classification deep learning (DL) model trained to inspect products for defects, comprising:
- a processor;
- and
- a computer-readable medium in communication with the processor and storing instructions that, when executed by the processor, cause the processor to control the system to perform:
- receiving a first image capturing a first product;
- providing the received image to an edge computer operating a classification DL model pre-trained to determine, based on a set of parameters, whether or not a product has an acceptable quality based on an image capturing the product, wherein the classification DL model is configured to perform:
- in response to the first image provided thereto, attempting to determine, based on the first image, whether or not the first product has the acceptable quality;
- determining that whether or not the first product has the acceptable quality cannot be determined using the first image;
- and
- in response to determining that whether or not the first product has the acceptable quality cannot be determined using the first image, determining that the first image contains an error image;
- in response to determining that the first image contains an error image, adjusting, based on the error image, the set of parameters of the classification DL model;
- and
- extracting, from the edge computer, the adjusted set of parameters of the classification DL model.
A non-transitory computer-readable medium comprising instructions that, when executed by a processor, cause the processor to control a system to perform functions of retraining a classification deep learning (DL) model trained to inspect products for defects, the functions including:
- receiving a first image capturing a first product;
- providing the received image to an edge computer operating a classification DL model pre-trained to determine, based on a set of parameters, whether or not a product has an acceptable quality based on an image capturing the product, wherein the classification DL model is configured to perform:
- in response to the first image provided thereto, attempting to determine, based on the first image, whether or not the first product has the acceptable quality;
- determining that whether or not the first product has the acceptable quality cannot be determined using the first image;
- and
- in response to determining that whether or not the first product has the acceptable quality cannot be determined using the first image, determining that the first image contains an error image;
- in response to determining that the first image contains an error image, adjusting, based on the error image, the set of parameters of the classification DL model;
- and
- extracting, from the edge computer, the adjusted set of parameters of the classification DL model.
Dependencies are inferred from the claim text.