Diabetic Retinopathy Multistage Classification Using EfficientNetB3 Model

Diabetic retinopathy (DR) is a disorder that continuously degrades human vision. It is recognized as major cause for the blindness in society accounting to around 40%. Growth of retinopathy is noticed in four different stages leading towards everlasting blindness. Prior detection of DR and suitable treatment can prevent consequence of permanent blindness. Present work discloses a means for segregating retinopathy into various stage of disorder using artificial intelligence. Designed model is made to understand set of Gaussian filtered retina images consisting of DR of different stages and tested for untrained image data to detect the phase of disorder. It is observed that the other experiments were merely determining whether or not the fundus image contained diabetic retinopathy. With balanced dataset EfficientNetB3 performs better with training accuracy of 100% and testing accuracy of 93.34%.