Deep-learning-based anomaly detection facilitates automated surface inspection for, e.g., detection and segmentation of defects. The technology is able to unerringly and independently localize deviations, i.e., defects of any type, on subsequent images. Model training only requires images of samples without defects. As opposed to other deep learning methods, no labeling effort is required. During inference, anomaly detection segments the areas of images that deviate significantly from the training images.
Global Anomaly Detection
In the image of an electronic board with minor differences such as missing or misplaced components, missing labels, or absent USB connection points can be observed.
Advantages of this approach:
- Detection of various types of anomalies in a global context
- Powerful algorithms
- No need for labeling
- Requires only good quality images.
