As featured in the August 23, 2021 Research and Innovation Newsletter: LanePainter, a GAN-based model for simultaneously classifying and enhancing low-quality lane marks, has been demonstrated to successfully detect low-quality lane markings, improving the performance of lane detection algorithms.
Brief Project Description: Lane detection is one of the popular research fields in computer vision and autonomous vehicles. However, the performance of lane mark detection algorithms under low-quality lane marking environments is rarely studied. In this paper, we study this problem and propose LanePainter, a GAN-based model for simultaneously classifying and enhancing the lane marks. In this work, we demonstrate that our model can successfully detect low-quality lane marks. And, the enhanced lane marks can improve the performance of existing lane detection algorithms on low-quality lane marks.
Poster: 
Fact Sheet Video Other Q&A
Presenter: Xiaohan Zhang, UVM