Abstarct: Image segmentation is one of the first steps in each machine vision system. Image segmentation problem could be surveyed in two branches: automatic image segmentation and interactive image segmentation. Here we will study graph cut baxsed interactive image segmentation. Success of graph cut baxsed segmentation is dependent suitable parameters selection. Usually each image has its own optimal set of parameters. In this research our goal is automatic evaluation of segmentation results. we will propose two systems using Adaboost and neural network classifiers for evaluating graph cut baxsed segmentation results. In each of the proposed systems ,classifiers is trained using training set. Then the trained classifiers can evaluate a segmented image. We extract 6 features from segmented images in training and testing stages. This features show homogeneous properties of segments and differences of boundary using intensity and texture. Experimental results shows an error rate of 1.7 % for Adaboost classifier and an error rate of 0.65 % for neural network classifier.