Brain tumor classification and segmentation in the medical field is still a challenging task. Because we cannot identify through our naked eyes. Even Though several algorithms and methods developed to segment the brain tumor still accuracy is needed .By the single level classification we may not obtain the accurate result. So we propose the CNN (Convolution Neural Network) classifier which contains several layers. The convolution neural network uses kernals.The classification here is used to find the brain tumors such as glioma,meningioma and pituitary .The classified image is segmented using the watershed algorithm which segments based on the intensity.The segmentation employs here is to find the size of the tumor.