Skip to content
Conference Open access

LaneVision: A VGG16-UNet Based Deep Learning Approach for Robust Road Lane Detection

Sep 2026 · 2026 IEEE International Conference on Advances in Computing Research On Science Engineering and Technology (ACROSET) · pp. 1-6 · 0 citations · 27 references

Abstract

Lane detection is a fundamental component of Advanced Driver Assistance Systems (ADAS) and autonomous driving technologies, playing a critical role in ensuring safe and reliable vehicle navigation. Conventional lane detection methods, including edge detection, Hough Transform, and curve fitting, often fail to perform effectively under challenging real-world conditions such as poor illumination, sharp road curves, heavy traffic, faded lane markings, occlusions, and adverse weather. To overcome these limitations, this paper proposes LaneVision, a deep learning-based lane detection framework that combines the feature extraction capability of VGG16 with the semantic segmentation strength of UGNet. The proposed model formulates lane detection as a pixel-wise segmentation task, enabling accurate localization of lane boundaries. Experiments were conducted using the TuSimple lane detection dataset containing 3,626 annotated highway images. To improve model generalization, data augmentation techniques, including image inversion and brightness adjustment, expanded the training dataset to 9,789 images. The encoder utilizes ImageNet-pretrained VGG16 convolutional layers with skip connections to a customized U-Net decoder incorporating batch normalization and dropout regularization. After 100 training epochs using the Adam optimizer and Dice Loss, the model achieved a validation accuracy of 98.04%, a Dice coefficient of 0.7788, 75.15% precision, and 80.87% recall. The trained model is deployed through a Flask-based REST API with a responsive HTML5 interface, providing real-time lane visualization. Experimental comparisons demonstrate that LaneVision outperforms traditional approaches and state-of-the-art models.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.