The precise delineation of retinal vasculature is vital for identifying and treating a variety of ocular disorders. This paper presents a novel enhanced Residual U-Net (Res-Unet) model intended to improve the precision and resilience of retinal vascular segmentation. The proposed model integrates residual connections into the U-Net architecture, a technique commonly employed to mitigate the gradient degradation issue in deep networks. This integration contributes to the model’s ability to retain the spatial information required for accurate vessel delineation. Additionally, to enhance the model’s performance, a new BCE logits loss function is employed. The model is trained and evaluated on the freely accessible DRIVE and STARE databases. The experimental findings reveal that it outperforms preceding techniques in terms of various performance indices such as accuracy, sensitivity, and specificity. The model's enhanced resilience and accuracy validate its potential as a dependable gauge for automated retinal image-based analysis, which will eventually aid in clinical applications.