Abstract:
The retina is a very important visionary element of our body. The elements of the retina are also very important and researchers all over the world are very keen to know & identify those elements. The current improvement in our technology and computer science gives us the proper power to segment a retinal image and also identify those elements of the retina simultaneously. The power of computing more and more gives us the effectiveness to use machine learning and computer vision and use more images to train the machine and get more accurate simultaneous accurate results. There are a lot of research happened in the field of machine learning and those are quite good and also accurate but in most of the case most of them try to identify one or two retinal elements in their research but we are proposing in our model that we can accurately identify three retinal elements and with more accuracy. There is two paper which of them identify three retinal elements like us but one of them is 1999th research and another one is 2017th and the newer one uses one CNN model to train the machine and one CNN model generating the output of those three elements so in that case, every element is not quite accurate.[1][2] We are proposing a new model with three CNN models which would be identifying every element distinctly so every CNN model would be trained for a single element of the retina so the resultant output would be much more accurate. We had calculated and tested that three CNN models can provide us at least 7 to 12 percent more accuracy than a single CNN model.[1] The three-element we are identified by image segmentation are optical disk, blood vessels and fovea. All the models are pre-trained and tested with the same dataset which was public and available.
