Algorithm 1 Steps to train the Custom CenterNet model

INPUT: TS, annotations OUTPUT: Localized RoI, CM, Classified lesion         TS–training set.         annotations–Position of DR and DME lesions in retinal images         Localized RoI–lesion location in the output         CM–CenterNet model with DenseNet-100 backbone.         Classified lesion–Class associated with each detected lesion. imageSize ← [x y] // Approximation of Bounding box     α ← AnchorsEstimation (TS, annotation) // CenterNet Model     CM ← DenseNet100Based CenterNet (imageSize, α)     [It, Is] ← division of samples into training and testing dataset // Training Unit of lesions Identification     For each image i from → It         Compute DenseNet100 keypoints → ts     End For     Training CM over ts, and calculate training time t_dense     η_ dense ← PreLesionPosition(ts)     Ap_ dense ← Evaluate_AP (DenseNet-100, η_ dense)     For each sample I from → Is           (a) extract features through trained framework € → βI           (b) [bounding_box, objectness_ score, output label] ← Predict (βI)           (c) show sample along with bounding_box, output label           (d) η ← [η bounding_box]     End For Ap_€ ← Evaluate framework € using η Output_class ← CM (Ap_€).