Deep learning streamlines material inspections
Multiphase analysis of composite materials is a typical industrial image analysis application using deep-learning technology. Left: original image of etched copper. Middle: image segmentation using conventional thresholding methods. Right: deep-learning image segmentation.
Waltham, Massachusetts – Olympus Stream image analysis software now leverages the power of artificial intelligence (AI) to bring next-generation image segmentation to industrial microscope inspections. Software version 2.5 adds Olympus’ TruAI deep-learning technology, enabling users to train neural networks to automatically segment and classify objects in microscope images for a range of material inspections. A trained network can be applied to future analyses for a similar application to maximize efficiency.
Image analysis is a critical part of many material science, industrial and quality assurance applications. However, image segmentation using conventional thresholding methods that depend on HSV or RGB color spaces can miss critical information or targets in samples. Olympus’ TruAI technology offers more accurate segmentation based on deep learning for a highly reproducible and robust analysis.
With the TruAI solution, users can easily train robust neural networks. An easy-to-use interface lets users efficiently label images and run trainings in batches. Networks can be configured with many input channels, trained to identify up to 16 classes, and imported or exported. The solution also offers options to review and edit training details.
The software update also gives all users access to Olympus’ workflow customization services. This team designs tailor-made Olympus Stream workflows to address user-specific application scenarios, challenges, and goals.
Olympus Stream v. 2.4 customers may use their existing license for a free update to software version 2.5.