Deep learning is emerging as one of the most widely used technologies in the modern world and is becoming increasingly important to medical institutions because it can automate disease diagnosis, allowing for the early detection and saving time. Diabetic retinopathy (DR) is a common illness that can cause vision loss, which can be effectively managed with early identification. This study suggests a novel approach that uses fundus images and deep learning to accurately diagnose DR. The pre-processing of these images includes the use of median filtering, contrast-limited adaptive histogram equalisation (CLAHE), and Canny edge detection for blood vessel segmentation. Using pre-trained ImageNet weights and transfer learning, InceptionNet leverages deep learning to perform feature extraction and categorisation. The aim of combining edge detection and deep learning approaches is to monitor and identify DR stages with high accuracy. Real-time fundus images are used in this system's experimentation, and the outcomes show how well the suggested strategy works to improve diagnostic tools for conditions that could cause blindness.