Detecting Phases in Steel Microstructure

Abstract

Detecting Phases in Steel Microstructure is one of the interesting problem in the field of computer vision. In this work, we discuss a pixel based classification approach. A classifier is only as good as the information you give it. On the other hand it may have a huge intrinsic disproportion the number of examples in each class, Which hinder the classification performance. There are many ways you can adjust how you’re representing your input data for learning of model. In this paper, we propose ensemble method based on outlier detection to comprehend better the data used as a part of learning of model , which is random forest and discuss it merits and demerits of other related method which we use.