In the framework of this research, we suggest integrated, trainable network that can go from beginning to finish, able to recognize road markers and use vanishing points for navigation in bad weather. We concentrate on wet and poorly light situations, which have received less attention owing to evident obstacles. Pictures captured during a downpour are one such example prone to poor lighting, and the occurence of lane markers is distorted by light reflection on wet roads. Color distortion occurs at night when there is minimal light. This has led to the lack of a standard dataset and only a tiny number of proven algorithms are weather-resistant. To address this shortcoming, we used a cutting-edge dataset labeled by VPGNet for lanes and road markings, which consists of approximately 20,000 photos taken in four distinct conditions: no rain, light rain, heavy rain, and night, supplemented by a smaller dataset of 2,000 images of three scenes: glare, night, and grainy captured from an donkeycar contraption. We create, test, and compare several iterations of the proposed networkwhilst reiterating the value of each individual role. The proposed method needs just one forward pass to identify lanes and road markers and predict vanishing points. Across a variety of experimental setups, our method has shown to be both accurate and reliable in real time. (25 fps).