DEEP CNN-BASED MULTI-CLASS RETINAL DISEASE DETECTION WITH LOW MEMORY CONSUMPTION

Abstract

 The examination fosters a memoryproficient convolutional neural network (CNN) model
to recognize and characterize retinal issues. U-Net
Division, utilized for retinal sickness characterization,
utilizes a great deal of memory and central processor.
The recommended model tends to these challenges.
This model is tried on EyeNet, a benchmark dataset
with 32 retinal sickness types. Trial results uncover
that the proposed model beats existing techniques in
memory the executives and exactness. Precision,
recall, and accuracy measures are assessed utilizing
various ages and step times. On the EyeNet dataset, the
proposed strategy groups multi-class retinal sicknesses
with phenomenal exactness. We utilized mobilenet,
densenet, and hybridways to deal with further develop
accuracy. MobileNet accomplished 97% accuracy,
Xception 100 percent, and A hybrid. For protected and
effective client connections, a Flask front-end with
validation is made.[38]