Unveiling Superior Lane Detection Techniques Through the Synergistic Fusion of Attention-Based Vision Transformers and Dense Convolutional Neural Networks

For navigating different kinds of surroundings, lane detection is an important technique in the field of autonomous vehicles. In this research work, a new methodology has been presented for accurately detecting the lanes by applying vision-based transformer and dense convolutional neural network. To capture the small attention details of different pixels, markings have been made for the understanding. Occlusion-based mask predictor has been applied to increase the robustness of the proposed model. Next, the patches are extracted and processed by Dense CNN networks to assess the fined grinned details of the extracted patches. Therefore, a much-improved classification accuracy has been achieved by the combination of a vision-based transformer and Dense CNN. This accuracy is obtained by taking the reference from Intersection over Union (IoU) as a reference. For getting high accuracy as well as precision, this proposed method would help in achieving this solution. As a result, the real-time problems in the safe transformation, this method could lead to a significant potential for enhancing the performance in real-time problems in the safe transformation.