Enhancing Blindness Detection using Deep Learning with MobileNet and SE Blocks
Priya Mittal, Bhisham Sharma, Dhirendra Prasad Yadav, Imed Ben Dhaou · 2025
Diabetic retinopathy (DR), a severe complication of diabetes, can lead to vision impairment or complete blindness, if not detected early. This paper presents a novel deep learning approach using an enhanced MobileNet architecture integrated with Squeeze-and-Excitation (SE) blocks to automatically detect DR at an early stage. Trained and validated on the APTOS 2019 Blindness Detection dataset, the proposed method significantly improves feature extraction and outperforms conventional techniques. The modified model achieved a precision of $79.21 \%$ and an accuracy of $89.67 \%$. These results demonstrate that the SE-MobileNet model substantially enhances diagnostic accuracy, establishing it as a promising tool for early DR detection. Additionally, its lightweight nature makes it suitable for deployment in cloud-based and real-time applications.