Spatial-UNet: Deep Learning-Based Lane Detection Using Fisheye Cameras for Autonomous Driving

Moujtahid, Salma; Benmokhtar, Rachid; Breheret, Amaury; Boukhdhir, Saif-Eddine

doi:10.1007/978-3-031-06430-2_48

Cited by 4 publications

(5 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We specifically chose to evaluate CLLD with U-Net because it is a common encoder-decoder architecture used in various methods that approach lane detection as a segmentationbased problem (Lee and Liu 2022;Zou et al 2019;Tran and Le 2019a;Moujtahid et al 2022). In addition, we tested our method using RESA (Zheng et al 2021), which is currently the state-of-the-art semantic segmentation lane detection method that is not based on the U-Net architecture.…”

Section: Resultsmentioning

confidence: 99%

“…Based on their applications, contrastive learning (CL) methods are divided into global feature (Chen et al 2020a;Bardes, Ponce, and LeCun 2022;Chen, Xie, and He 2021;Wang et al 2021;Mitrovic et al 2020;Bardes, Ponce, and LeCun 2021;Mitrovic et al 2020;He et al 2020;Henaff 2020;Tian, Krishnan, and Isola 2020) and local feature (Wang et al 2021;Bardes, Ponce, and LeCun 2022;Xie et al 2021;Zhao et al 2021;Xiao et al 2021) CLs.…”

Section: Contrastive Learningmentioning

confidence: 99%

“…A robust lane detection model must capture spatial information to detect poorly visible parts of lanes using other areas that are more visible (Pan et al 2018). Many novel lane detection approaches employ pixellevel image segmentation techniques (Zheng et al 2021;Pan et al 2018;Moujtahid et al 2022) to enhance the precision of lane detection. In this group of methods, each pixel is labeled as belonging either to a lane or the background.…”

Section: Introductionmentioning

confidence: 99%

“…We assess CLLD on U-Net (Ronneberger, Fischer, and Brox 2015), a popular encoder-decoder which widely utilized in lane detection algorithms (Lee and Liu 2022;Zou et al 2019;Tran and Le 2019a;Moujtahid et al 2022). In addition to U-Net, we evaluate the proposed method on Recurrent Feature-Shift Aggregator for Lane Detection (RESA) (Zheng et al 2021) and Cross Layer Refinement Network for Lane Detection (CLRNet) (Zheng et al 2022) as state-of-the-art segmentation-based and anchor-based lane detection methods, respectively.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

3DLaneNAS: Neural Architecture Search for Accurate and Light-Weight 3D Lane Detection

Zoljodi

Loni

Abadijou

et al. 2022

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Detecting road lanes is challenging due to intricate markings vulnerable to unfavorable conditions. Lane markings have strong shape priors, but their visibility is easily compromised. Factors like lighting, weather, vehicles, pedestrians, and aging colors challenge the detection. A large amount of data is required to train a lane detection approach that can withstand natural variations caused by low visibility. This is because there are numerous lane shapes and natural variations that exist. Our solution, Contrastive Learning for Lane Detection via cross-similarity (CLLD), is a self-supervised learning method that tackles this challenge by enhancing lane detection models' resilience to real-world conditions that cause lane low visibility. CLLD is a novel multitask contrastive learning that trains lane detection approaches to detect lane markings even in low visible situations by integrating local feature contrastive learning (CL) with our new proposed operation cross-similarity. Local feature CL focuses on extracting features for small image parts, which is necessary to localize lane segments, while cross-similarity captures global features to detect obscured lane segments using their surrounding. We enhance cross-similarity by randomly masking parts of input images for augmentation. Evaluated on benchmark datasets, CLLD outperforms state-of-the-art contrastive learning, especially in visibility-impairing conditions like shadows. Compared to supervised learning, CLLD excels in scenarios such as shadows and crowded scenes.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Contrastive Learningmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

3DLaneNAS: Neural Architecture Search for Accurate and Light-Weight 3D Lane Detection

Zoljodi

Loni

Abadijou

et al. 2022

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…Lane detection is crucial for Advanced Driver Assistance Systems (ADAS). Currently, camera-based lane detection [1][2][3] is the most popular lane detection method. However, camera is easily influenced by lighting and has low accuracy under poor lighting conditions.…”

Section: Introductionmentioning

confidence: 99%

A detection frame selection method for distinguishing between solid and dashed lane lines using LiDAR

Wang,

Ding,

Guan

et al. 2024

Seventh International Conference on Traffic Engineering and Transportation System (ICTETS 2023)

View full text Add to dashboard Cite

The 16-line LiDAR-based lane detection method is inexpensive and widely used, but it has the following problems. The point cloud distribution is sparse and unable to distinguish between solid and dashed lines when the number of detection frames is too small; when the number is too large, the data processing efficiency is low. Therefore, it is necessary to find the minimum number of detection frames required to distinguish between solid and dashed lines. To address this problem, this paper established a model to calculate the minimum number considering the factors such as lane, LiDAR installation height and vehicle speed. Based on the simulation results, this paper summarized a calculation method of the minimum number suitable for a variety of common working conditions.

show abstract

Recent Trends in Autonomous Vehicle Validation Ensuring Road Safety with Emphasis on Learning Algorithms

Abraham

Nagavarapu

Prasad

et al. 2022

2022 17th International Conference on Control, Automation, Robotics and Vision (ICARCV)

View full text Add to dashboard Cite

Spatial-UNet: Deep Learning-Based Lane Detection Using Fisheye Cameras for Autonomous Driving

Cited by 4 publications

References 13 publications

3DLaneNAS: Neural Architecture Search for Accurate and Light-Weight 3D Lane Detection

3DLaneNAS: Neural Architecture Search for Accurate and Light-Weight 3D Lane Detection

A detection frame selection method for distinguishing between solid and dashed lane lines using LiDAR

Recent Trends in Autonomous Vehicle Validation Ensuring Road Safety with Emphasis on Learning Algorithms

Contact Info

Product

Resources

About