Various Machine learning and computer vision techniques are available for image processing.
In the last few years, the field of machine learning has made tremendous progress on addressing the difficult problems of Computer Vision and NLP. In particular, we've found that a kind of model called a deep convolutional neural networkcan achieve reasonable performance on hard visual recognition tasks -- matching or exceeding human performance in some domains.
This is an Image classifier based on CNN which is trained on ImageNet dataset and we are retraining it by using the approach of transfer learning and given the labelled data, it produces accuracy above 95% most of the time.
Fault and anomaly is detected on the simplified problem of metal surfaces.
One of the type is "no fault" type which is generated taking a cleaned metal surface and synthesising new images from it and other faults involve the "scratches", "pitted surface", "crazing", "inclusion", "rolled-in" and "patches".
Credits:
Algorithm : GoogleNet
Dataset : NEU surface defect dataset