Abstract:
Globally this second blinding condition which affects millions of people is the reason for the advocacy of early diagnosis
and treatment needed to stop these conditions from gettiong worse in the future Deep Learning as a Sustainable Technology to
Detect Retinal Diseases: In this article, we touch on the elements of retinal disease diagnosis using the deep learning models. The
tool, which adopts an ensemble of optimized state-of-the-art neural network architectures including MobileNetV2, ResNet50,
InceptionV3, and DenseNet, will undergo a detailed evaluation of their respective performance with classifying retinal images. The
strategy set behind our preprocessing steps, resize images, convert grayscale to RGB and extensive training cycles gives rise in the
evolution of a top model. Interestingly all the results showcase ResNet50 the best persoiling which produces accuracy of 0.89;
consequently, setting a new mark on retinal scan analysis. This research yields up a valular aspect of early retinal disease detection;
we end up having an increased chance of conducting a precise diagnosis and improved patient outcomes.
Using retinal fundus pictures, ophthalmologists may diagnose retinal issues with great precision. Early detection helps avoid
blindness and increase the likelihood of a cure. Medical professionals can diagnose retinal fundus images to help with conditions
including diabetic retinopathy and retinitis pigmentosa. Machine learning research has recently concentrated on using feature
extraction and image classification to diagnose conditions such as diabetic retinopathy. Our objective in this work is to automatically
identify, without explicit segmentation or feature extraction, photos with retinal abnormalities from those of the healthy. Instead, we
automatically categorise every retinal fundus image as healthy or sick using a deep learning algorithm.
