Our eyes are trained to quickly determine if a surface is uniform or if any anomalous part can be found on it. Recent developments on deep learning and visual anomaly detection allow for mimicking the process that is performed by human eyes using smart computer programs.
M2L Visual Anomaly Detection (VAD) identifies the part of the image that are inconsistent with the rest of it. It is an unsupervised technique based on latest advances in machine learning such as deep learning. Please try it here.