Abstract: Current transformer-based 3D lane detection methods typically use instance activation maps (IAM) and point-to-point loss to achieve small geometric deviations of lanes. However, these ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results