Unlike Tesla, emerging domestic automakers typically employ a fusion perception approach for intelligent driving, primarily centered around LiDAR and complemented by millimeter-wave radar, high-definition cameras, and other technologies. However, with the advancement of Tesla's FSD and the adoption of end-to-end technology, the effectiveness of a pure vision-based solution in high-level intelligent driving has garnered significant attention. This approach not only offers substantial cost advantages but also continuously enhances algorithm iteration and practical application capabilities.