Seshikala Code.zip

This paper presented the HODLNet deep neural network, an effective optimization-based system for evaluating Non-Proliferative Diabetic Retinopathy severity levels. The noise is removed and the image is made grayscale during the preprocessing stage. Segmenting and removing the blood vessels and optic disc, which improves the area with lesions and eliminates the spurious regions. A feature vector is created by first extracting the features and then combining them for the desired classification. The stages of the input fundus image have been categorized as normal, mild, moderate, and severe using the improved ShuffleNet v2 model based on the features of the lesions.