An Explainable Transfer Learning Framework for Early Detection of Diabetic Retinopathy Using Deep Learning

The major contributions of this work are summarized as
follows:
• Acomprehensive comparison of six state-of-the-art trans
fer learning architectures for multiclass diabetic retinopa
thy classification.
• Development of an ensemble framework combining
MobileNetV3-Large and ConvNeXt Tiny to improve pre
diction robustness.
• Integration of Grad-CAM visualization for explainable
diabetic retinopathy diagnosis.
• Extensive experimental evaluation using the publicly
available APTOS 2019 Blindness Detection dataset.