Enhanced Image Segmentation Using Convolutional Recurrent Neural Networks

Abstract

Accurate Image Segmentation is vital for distinguishing the elements and errors in several Convolutional Neural Networks(CNN).It has become the state of art automatic segmentation strategies. Still, the totally automatic results are required to be refined to become correct and strong enough so as to assist in clinical use. we tend to here propose a deep learning based interactive segmentation technique that improves the results obtained by the automated CNN and also the mis-segmented elements are re-segmented by Recurrent Neural Networks(RNN) that cut back the user interaction throughout the refinement of upper accuracy. we tend to propose to mix each the CNN and RNN that will increase the accuracy of segmentation by reducing the user interaction. Experimental results shows that our technique can come through giant improvement from the automated (Convolutional repeated Neural Networks) CRNN and it'll acquire comparable even high accuracy.