In the field of materials science, phase diagram construction is often a laborious process because of the time taken for a human to sift through a large number of images. To mitigate this, machine learning approaches have been used by researchers to automate the construction of these phase diagrams using both supervised and unsupervised learning. In this dissertation, I have outlined a step-wise lego-like approach using both unsupervised learning and image processing approaches to automate the construction of image based phase diagrams and have illustrated that with the construction of three polymer phase diagrams.