Overview

Automated inspection and defect detection are critical for high throughput quality control in production systems. They are widely adopted in many industries for detection of flaws on different kinds of manufactured surfaces such as metallic rails, semiconductor wafers, contact lenses and so on. Recent developments in deep learning have brought revolutionary new tools to automate visual inspection tasks with unprecedented accuracy and robustness. New methods allow you to find arbitrary defects without the need for failure data during training.

In this session we will discuss state-of-the-art approaches for visual inspection and present multiple case studies on how these approaches have been applied in industry. We will cover basic classification problems as well as advanced algorithms like autoencoders, semantic segmentation, generational adversarial networks and one-class learning. Through a couple of hands-on examples you will see how it can be implemented in MATLAB. We will also discuss methods to tackle common pitfalls in the development of such models, and challenges presented in validating and operationalizing a deep network.

Highlights:

• Preprocessing and augmenting image data
• Overview of methods available for automated visual inspection
• How to train a deep neural network in MATLAB
• Managing experiments and tuning training parameters
• Automatically generate code and deploy to embedded targets
• Validating and operationalizing models