Quality control of products in industry is often still done manually by experienced operators. This is often found to be inefficient, time consuming and error prone. Therefore, automated approaches have been introduced to assist the operator in this task. Generally, automated inspection systems consist of the following modules: image acquisition, image processing, feature extraction and decision making. The automated approach can be based on classic computer vision methods, but these are often based on fixed rules and therefore not easily adaptable to new data or scenarios.

Industrial interest in machine learning applications has exploded since 2005 and the growing trend is evident. Furthermore, neural networks (especially deep architectures) have been important from 2012 onwards, thanks to their versatility and excellent quality of results.