Diabetic retinopathy (DR) is a leading cause of blindness globally, requiring timely detection and classification to prevent vision loss. Deep learning techniques offer significant potential for automating DR detection by analyzing retinal fundus images with high precision. This paper proposes a RetinoNet model that consists of MobileNetV3, Convolutional Block Attention Module (CBAM), Atrous Spatial Pyramid Pooling (ASPP), and Feature Pyramid Network (FPN). MobileNetV3 provides a lightweight and efficient foundation for feature extraction, while CBAM emphasizes critical spatial and channel information, enabling the detection of subtle retinal abnormalities. ASPP captures multi-scale contextu... More >