2021 IEEE International Conference on Image Processing

19-22 September 2021 • Anchorage, Alaska, USA

Imaging Without Borders

2021 IEEE International Conference on Image Processing

19-22 September 2021 • Anchorage, Alaska, USA

Login Paper Search My Schedule Paper Index Help

My ICIP 2021 Schedule

Note: Your custom schedule will not be saved unless you create a new account or login to an existing account.
  1. Create a login based on your email (takes less than one minute)
  2. Perform 'Paper Search'
  3. Select papers that you desire to save in your personalized schedule
  4. Click on 'My Schedule' to see the current list of selected papers
  5. Click on 'Printable Version' to create a separate window suitable for printing (the header and menu will appear, but will not actually print)

Paper Detail

Paper IDSS-AVV.8
Paper Title ROBUST MONOCULAR 3D LANE DETECTION WITH DUAL ATTENTION
Authors Yujie Jin, Xiangxuan Ren, Shanghai Jiao Tong University, China; Fengxiang Chen, Tongji University, China; Weidong Zhang, Shanghai Jiao Tong University, China
SessionSS-AVV: Special Session: Autonomous Vehicle Vision
LocationArea A
Session Time:Monday, 20 September, 13:30 - 15:00
Presentation Time:Monday, 20 September, 13:30 - 15:00
Presentation Poster
Topic Special Sessions: Autonomous Vehicle Vision
IEEE Xplore Open Preview  Click here to view in IEEE Xplore
Abstract Getting an accurate estimation of three-dimensional position of the driveable lane is crucial for autonomous driving. In this work, we introduce a novel attention module called Dual Attention (DA) which enables the model to perform robustly and accurately under complicated enviromental conditions. More specifically, the attention mechanism adopts a two- pathway correlated attention method to produce additional features and aggregate globle information. We demonstrate the effectiveness of our method by following and extending recently proposed state-of-the-art 3D lane marking detection methods. Moreover, we use a novel linear-interpolation loss to precisely fit the lane marking. Extensive conducted experiments demonstrate that our methods outperform baseline methods on Apollo synthetic 3D dataset.