Year-end Sale % OFF 
Abstract
Thanks to recent advancements in image processing and deep learning techniques, visual surface inspection in production lines has become an automated process as long as all the defects are visible in a single or a few images. However, it is often necessary to inspect parts under many different illumination conditions to capture all the defects. Training deep networks to perform this task requires large quantities of annotated data, which are rarely available and cumbersome to obtain. To alleviate this problem, we devised an original augmentation approach that, given a small image collection, generates rotated versions of the images while preserving illumination effects, something that random rotations cannot do. We introduce three real multi-illumination datasets, on which we demonstrate the effectiveness of our illumination preserving rotation approach. Training deep neural architectures with our approach delivers a performance increase of up to 51% in terms of AuPRC score over using standard rotations to perform data augmentation.
Highlights
Defect detection has long been central to a wide range of industrial vision systems that are required to provide reliable, high-speed, and cost-effective quality checks at frequent intervals
We evaluate the performance of different techniques in terms of the area under the precision-recall curve (AuPRC), known as the average precision
To reduce the effect of random initialization on performance, all AuPRC values reported are computed as the mean over two independent runs
Summary
Defect detection has long been central to a wide range of industrial vision systems that are required to provide reliable, high-speed, and cost-effective quality checks at frequent intervals. Acquiring several images while varying the light orientation and intensity without moving the cam- Addressing these limitations with a generic deep-learning approach is not a simple matter of running a standard model on the image stacks. This is essentially because capturing the defects requires using a large number of images, which contain much redundant information. Training a deep network given only limited amounts of annotated data, which is the norm due to the scarcity of defective samples, results in overfitting This can be addressed by various data augmentation techniques, including applying random photometric and geometric transformations such as brightness changes, translations, or rotations to the images. Applying random rotations to the image stacks needlessly disrupts the connection between
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
