• Manojkumar S B
• Dr. H S Sheshadri

Here we proposed a deep learning method for detecting diabetic retinopathy (DR). DR is one of the common for prolong standing of diabetics later it leads to vision loss. In our early papers we worked with MATLAB software and applied various algorithms like K-means cluster algorithm, SVM and Random forest algorithm. Now we proposed transfer learning process (inception v3) method of tensor flow software by adding the region proposal networks (RPN) after the global averaging pooling layer of the convolutional networks (CNN) for classification of diabetic retinopathy and Faster R-CNN for edudates identification and extraction. The proposed model can localize the regions of a retina image to show the specific region of interest in terms of its severity level. Our opinion is this deep learning method is highly desired for diabetic retinopathy detection because in practice, users are not only interested with high prediction performance, but also keen to understand the insights of DR detection and transfer learning model works. In the experiments conducted on a large scale of retina image dataset collected from kaggle, a number of pre-trained models are available which perform object classification from an image of the object which is given as an input. However, we can retrain the pre-trained models for the set of classes of interest and in this work; one such model has been retained to measure the severity of Diabetic Retinopathy in eye images (data has been provided by eyepacs as color fundus images) on the scale of 0-5 we show that the proposed faster R-CNN model can achieve high performance on DR detection compared with the state-of-the-art while achieving the merits of providing the RPN to highlight the salient regions of the input image and we achieved 83.3% raw image accuracy A convolutional neural network classifier engineered from Inception V3 network which is trained for ImageNet, for 5-class severity classification performed best with an accuracy of 83.3%.

Diabetic Retinopathy, Faster R-CNN, Transfer Learning, RPN.

"Diabetic Retinopathy Classification Using Transfer Learning and Exudates Detection Using Faster–RCNN", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.6, Issue 5, page no.29-34, May-2019, Available :http://www.jetir.org/papers/JETIR1905L05.pdf

"Diabetic Retinopathy Classification Using Transfer Learning and Exudates Detection Using Faster–RCNN", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.6, Issue 5, page no. pp29-34, May-2019, Available at : http://www.jetir.org/papers/JETIR1905L05.pdf