VPGNet: Vanishing Point Guided Network for Lane and Road Marking Detection and Recognition

By Seokju Lee, Junsik Kim, Jae Shin Yoon, Seunghak Shin, Oleksandr Bailo, Namil Kim, Taeyeop Lee, H. Hong, Seung-Hoon Han, In-So Kweon. IEEE International Conference on Computer Vision, 2017

448 citations recorded by the source corpus.

In this paper, we propose a unified end-to-end trainable multi-task network that jointly handles lane and road marking detection and recognition that is guided by a vanishing point under adverse weather conditions. We tackle rainy and low illumination conditions, which have not been extensively studied until now due to clear challenges. For example, images taken under rainy days are subject to low illumination, while wet roads cause light reflection and distort the appearance of lane and road markings. At night, color distortion occurs under limited illumination. As a result, no benchmark dataset exists and only a few developed algorithms work under poor weather conditions. To address this shortcoming, we build up a lane and road marking benchmark which consists of about 20,000 images with 17 lane and road marking classes under four different scenarios: no rain, rain, heavy rain, and night. We train and evaluate several versions of the proposed multi-task network and validate the importance of each task. The resulting approach, VPGNet, can detect and classify lanes and road markings, and predict a vanishing point with a single forward pass. Experimental results show that our approach achieves high accuracy and robustness under various conditions in realtime (20 fps). The benchmark and the VPGNet model will be publicly available

Subfields: Automotive Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/22d046e926d3ed12d00063b8668cb11fed26e1f1

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Automotive Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

CondLaneNet: a Top-to-down Lane Detection Framework Based on Conditional Convolution (2021, methodology): “Segmentation-based methods [28, 12, 27, 19, 21, 6] are most common and have achieved impressive performance.”
Towards End-to-End Lane Detection: an Instance Segmentation Approach (2018, background): “[21] show how a multi-task network can jointly handle lane and road marking detection and recognition under adverse weather and low illumination conditions.”
BDD100K: A Diverse Driving Video Database with Scalable Annotation Tooling (2018, background): “Caltech Lanes Dataset [2] - 1,224 4 1 1 2 Road Marking Dataset [21] - 1,443 29 2 3 10 KITTI-ROAD [11] 289 579 1 1 2 VPGNet [13] 14,783 21,097 4 2 17 Ours 70,000 100,000 100,000 6 3 11 Table 2: Comparisons with other lane marking datasets.”
Robust Lane Detection From Continuous Driving Scenes Using Deep Neural Networks (2019, unknown): “On the purpose of multi-task learning, a VPGNet was proposed in [9], which has a shared feature extractor and four similar branch layers.”
Learning Lightweight Lane Detection CNNs by Self Attention Distillation (2019, methodology): “Contemporary algorithms [5, 8, 14, 16] typically adopt”
VectorMapNet: End-to-end Vectorized HD Map Learning (2022, unknown): “These methods typically involve handcrafted elements such as vanishing points [34], polynomial curves [35], line segments [36], and Bézier curves [37] to model proposals.”
Ultra Fast Structure-aware Deep Lane Detection (2020, background): “For instance, VPGNet [17] proposes a multi-task network guided by vanishing points for lane and road marking detection.”
Self-Driving Cars: A Survey (2019, methodology): “cently, several methods (Lee et al., 2017; Bailo et al., 2017; Ahmad et al., 2017) employed Maximally Stable Extremal Regions (MSER) for detecting regions of interest (i.e., regions that are likely to contain a road marking) and convolutional networks for recognizing road markings. Bailo et al. (201”
Line-CNN: End-to-End Traffic Line Detection With Line Proposal Unit (2020, methodology): “Note that here we do not include VPGNet [10] since it heavily relies on the birds-eye view by inverse perspective mapping (IPM) [54] and Vanishing Point (VP) to locate the traffic lines, neither of which are available in MIKKI.”
Computing Systems for Autonomous Driving: State of the Art and Challenges (2020, background): “For instance, VPGNet [97] proposes a multitask network for lane marking detection.”
BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learning (2018, background): “The most recent work, VPGNet [18], consists of about 20,000 images taken during three weeks of driving in Seoul.”