Home

Awesome

Awesome-Lane-Detection

Awesome

arXiv

overview

This repository is used for recording and tracking recent monocular lane detection methods, as a supplement to our survey paper:

Title: Monocular Lane Detection Based on Deep Learning: A Survey <br> Authors: Xin He, Haiyun Guo, Kuan Zhu, Bingke Zhu, Xu Zhao, Jianwu Fang, Jinqiao Wang<br> arXiv preprint arXiv:2411.16316<br>

This repository will be constantly updated. If you find any work missing or have any suggestions (papers, implementations and other resources), feel free to pull requests. We will add the missing papers to this repo ASAP.

Overview

overview

Summary of Contents

Methods of 2D Lane Detection

overview

(a)Segmentation-based methods (two-stage), which complete lane recognition and instance discrimination in a certain order, and leverage mask, grids or keypoints to model lanes.

(b)Object detection-based methods (one-stage), which can directly perform instance discrimination and localization concurrently, and leverage line anchor or parameter curve to model lanes.

Segmentation-based Methods

"↑" represents a bottom-up approach, usually distinguishing all lane foreground first, and then obtaining each lane instance through heuristic post-processing.

"↓" represents a top-down approach, which first identifies all instances and then predicts the specific location of each lane. "Max Lanes" means predefined maximum number of lanes in advance, which can be referred to in SCNN. “Dynamic kernels” denotes the method of predicting dynamic instance kernels to distinguish different instances, which can be referred to the classic instance segmentation methods Condinst and SOLOv2.

"None" indicates that the paper does not clearly indicate how to distinguish instances of different lanes.

MethodsVenueTitlePaper/CodeInstance DiscriminationLane Modeling
VPGNetICCV 2017VPGNet: Vanishing Point Guided Network for Lane and Road Marking Detection and RecognitionPaper/CodeMask
LaneNetIV 2018Towards End-to-End Lane Detection: an Instance Segmentation ApproachPaper/CodeMask
LaneNetArxiv 2018LaneNet: Real-Time Lane Detection Networks for Autonomous DrivingPaper/CodeMask
SCNNAAAI 2018Spatial As Deep: Spatial CNN for Traffic Scene UnderstandingPaper/Code↓ (Max Lanes)Mask
LMDDSP 2018Efficient Road Lane Marking Detection with Deep LearningPaper/CodeMask
EL-GANECCVW 2018EL-GAN: Embedding Loss Driven Generative Adversarial Networks for Lane DetectionPaper/Code↓ (Max Lanes)Mask
Chougule et al.ECCVW 2018Reliable multilane detection and classification by utilizing CNN as a regression networkPaper/Code↓ (Max Lanes)Keypoints
SADICCV 2019Learning Lightweight Lane Detection CNNs by Self Attention DistillationPaper/Code↓(Max Lanes)Mask
FastDrawICCV 2019FastDraw: Addressing the Long Tail of Lane Detection by Adapting a Sequential Prediction NetworkPaper/CodeMask
IntRA-KDCVPR 2020Inter-Region Affinity Distillation for Road Marking SegmentationPaper/Code↓ (Max Lanes)Mask
SALMNetTITS 2020SALMNet: A Structure-Aware Lane Marking Detection NetworkPaper/CodeNoneMask
Ripple-GANTITS 2020Ripple-GAN: Lane Line Detection With Ripple Lane Line Detection Network and Wasserstein GANPaper/CodeNoneMask
PINetTITS 2020Key Points Estimation and Point Instance Segmentation Approach for Lane DetectionPaper/CodeKeypoints
E2E-LMDCVPRW 2020End-to-End Lane Marker Detection via Row-wise ClassificationPaper/Code↓(Max Lanes)Keypoints
UFLDECCV 2020Ultra Fast Structure-aware Deep Lane DetectionPaper/Code↓(Max Lanes)Grids
RESAAAAI 2021RESA: Recurrent Feature-Shift Aggregator for Lane DetectionPaper/Code↓(Max Lanes)Mask
FOLOLaneCVPR 2021Focus on Local: Detecting Lane Marker from Bottom Up via Key PointPaper/CodeKeypoints
CondLaneNetICCV 2021CondLaneNet: a Top-to-down Lane Detection Framework Based on Conditional ConvolutionPaper/Code↓(Dynamic Kernels)Grids
LaneAFRAL 2021LaneAF: Robust Multi-Lane Detection with Affinity FieldsPaper/CodeMask
GANetCVPR 2022A Keypoint-based Global Association Network for Lane DetectionPaper/CodeKeypoints
UFLDv2TPAMI 2022Ultra Fast Deep Lane Detection with Hybrid Anchor Driven Ordinal ClassificationPaper/Code↓(Max Lanes)Grids
RCLaneECCV 2022RCLane: Relay Chain Prediction for Lane DetectionPaper/CodeKeypoints
CANetICASSP 2023CANet: Curved Guide Line Network with Adaptive Decoder for Lane DetectionPaper/Code↓(Dynamic Kernels)Grids
PriorLaneICRA 2023PriorLane: A Prior Knowledge Enhanced Lane Detection Approach Based on TransformerPaper/Code↓(Max Lanes)Mask
CondLSTRICCV 2023Generating Dynamic Kernels via Transformers for Lane DetectionPaper/Code↓(Dynamic Kernels)Keypoints
LanePtrNetArxiv 2024LanePtrNet: Revisiting Lane Detection as Point Voting and Grouping on CurvesPaper/CodeKeypoints

Object Detection-based Methods

MethodsVenueTitlePaper/CodeLane Modeling
Line-CNNTITS 2019Line-CNN: End-to-End Traffic Line Detection With Line Proposal UnitPaper/CodeLine Anchor
PointLaneNetIV 2019PointLaneNet: Efficient end-to-end CNNs for Accurate Real-Time Lane DetectionPaper/CodeLine Anchor
CurveLane-NASECCV 2020CurveLane-NAS: Unifying Lane-Sensitive Architecture Search and Adaptive Point BlendingPaper/CodeLine Anchor
PolyLaneNetICPR 2020PolyLaneNet: Lane Estimation via Deep Polynomial RegressionPaper/CodePolynomial
LaneATTCVPR 2021Keep your Eyes on the Lane: Real-time Attention-guided Lane DetectionPaper/CodeLine Anchor
SGNetIJCAI 2021Structure Guided Lane DetectionPaper/CodeLine Anchor
LSTRWACV 2021End-to-end Lane Shape Prediction with TransformersPaper/CodePolynomial
LaneFormerAAAI 2022Laneformer: Object-aware Row-Column Transformers for Lane DetectionPaper/CodeLine Anchor
EigenlanesCVPR 2022Eigenlanes: Data-Driven Lane Descriptors for Structurally Diverse LanesPaper/CodeLine Anchor
CLRNetCVPR 2022CLRNet: Cross Layer Refinement Network for Lane DetectionPaper/CodeLine Anchor
BézierLaneNetCVPR 2022Rethinking Efficient Lane Detection via Curve ModelingPaper/CodeBézier Curve
O2SFormerArxiv 2023End-to-End Lane detection with One-to-Several TransformerPaper/CodeLine Anchor
PGA-NetTITS 2023PGA-Net: Polynomial Global Attention Network With Mean Curvature Loss for Lane DetectionPaper/CodePolynomial
ADNetICCV 2023ADNet: Lane Shape Prediction via Anchor DecompositionPaper/CodeLine Anchor
CLRmatchNetArxiv 2023CLRmatchNet: Enhancing Curved Lane Detection with Deep Matching ProcessPaper/CodeLine Anchor
CLRerNetWACV 2024CLRerNet: Improving Confidence of Lane Detection with LaneIoUPaper/CodeLine Anchor
SRLaneAAAI 2024Sketch and Refine: Towards Fast and Accurate Lane DetectionPaper/CodeLine Anchor
HGLNetAAAI 2024A Hybrid Global-Local Perception Network for Lane DetectionPaper/CodeLine Anchor
GSENetAAAI 2024GSENet: Global Semantic Enhancement Network for Lane DetectionPaper/CodeLine Anchor
LDTRCVM 2024LDTR: Transformer-based Lane Detection with Anchor-chain RepresentationPaper/CodeAnchor-chain(uniform point)
Sparse LaneformerArxiv 2024Sparse LaneformerPaper/CodeLine Anchor

Methods of 3D Lane Detection

overview

(a)BEV-based Methods. The core is the view transformation from FV features to BEV features, including IPM and learning approach.

(b)BEV-free Methods. There are two branches: one is to project 2D lanes into 3D space based on depth estimation results, and the other is to directly model 3D lanes and project them back into FV for interaction and alignment.

BEV-based Methods

MethodsVenueTitlePaper/CodeView TransformationTask ParadigmLane Modeling
3D-LaneNetICCV 20193D-LaneNet: End-to-End 3D Multiple Lane DetectionPaper/CodeIPMODetLine Anchor
Gen-LaneNetECCV 2020Gen-LaneNet: A Generalized and Scalable Approach for 3D Lane DetectionPaper/CodeIPMODetLine Anchor
Efrat et al.Arxiv 2023Semi-Local 3D Lane Detection and Uncertainty EstimationPaper/CodeIPMSeg - ↑Keypoints
3D-LaneNet+Arxiv 20203D-LaneNet+: Anchor Free Lane Detection using a Semi-Local RepresentationPaper/CodeIPMSeg - ↑Keypoints
CLGoAAAI 2022Learning to Predict 3D Lane Shape and Camera Pose from a Single Image via Geometry ConstraintsPaper/CodeIPMODetPolynomial
Li et al.CVPRW 2022Reconstruct from top view: A 3d lane detection approach based on geometry structure priorPaper/CodeIPMODetLine Anchor
PersFormerECCV 2022PersFormer: 3D Lane Detection via Perspective Transformer and the OpenLane BenchmarkPaper/CodePersFormerODetLine Anchor
BEV-LaneDetCVPR 2023BEV-LaneDet: a Simple and Effective 3D Lane Detection BaselinePaper/CodeVPNSeg - ↑Keypoints
3D-SplineNetWACV 20233D-SpLineNet: 3D Traffic Line Detection using Parametric Spline RepresentationsPaper/CodeIPMODetB-Spline Curve
Chen et al.Arxiv 2023An Efficient Transformer for Simultaneous Learning of BEV and Lane Representations in 3D Lane DetectionPaper/CodeDecomposed TransformerSeg - ↓(Dynamic Kernels)Keypoints
Yao et al.ICCV 2023Sparse Point Guided 3D Lane DetectionPaper/CodePersFormerODetLine Anchor
GroupLaneArxiv 2023GroupLane: End-to-End 3D Lane Detection with Channel-wise GroupingPaper/CodeLSSSeg - ↓(Group Conv)Grids
LaneCPPCVPR 2024LaneCPP: Continuous 3D Lane Detection using Physical PriorsPaper)/CodeLSSODetB-Spline Curve

BEV-free Methods

Although PVALane constructs BEV feature, BEV feature are only used to assist in enhancing the 3D lane detection effect, rather than being a necessary component of the network like the BEV-based method.

MethodsVenueTitlePaper/CodeTask ParadigmLane Modeling
SALADCVPR 2022ONCE-3DLanes: Building Monocular 3D Lane DetectionPaper/CodeSeg - ↑Mask
CurveFormerICRA 2023CurveFormer: 3D Lane Detection by Curve Propagation with Curve Queries and AttentionPaper/CodeODet3D Line Anchor
Anchor3DLaneCVPR 2023Anchor3DLane: Learning to Regress 3D Anchors for Monocular 3D Lane DetectionPaper/CodeODet3D Line Anchor
LATRICCV 2023LATR: 3D Lane Detection from Monocular Images with TransformerPaper/CodeODet3D Line Anchor
DecoupleLaneArxiv 2023Decoupling the Curve Modeling and Pavement Regression for Lane DetectionPaper/CodeODet3D Polynomial
PVALaneAAAI 2024PVALane: Prior-Guided 3D Lane Detection with View-Agnostic FeaturePaper/CodeODet3D Line Anchor
BézierFormerICME 2024BézierFormer: A Unified Architecture for 2D and 3D Lane DetectionPaper/CodeODet3D Bézier Curve

Extended Works of Lane Detection

There are also some works that have received widespread attention in recent years, which are closely related to lane detection. In terms of task flow, they can be regarded as an upgrade on monocular image lane detection.

overview

Multi-task Perception

Methods2D/3D LaneVenueTitlePaper/Code
DLT-Net2DTITS 2019DLT-Net: Joint Detection of Drivable Areas, Lane Lines, and Traffic ObjectsPaper/Code
YOLOP2DMIR 2022YOLOP: You Only Look Once for Panoptic Driving PerceptionPaper/Code
HybridNets2DArxiv 2022HybridNets: End-to-End Perception NetworkPaper/Code
YOLOPv22DArxiv 2022YOLOPv2: Better, Faster, Stronger for Panoptic Driving PerceptionPaper/Code
TwinLiteNet2DMAPR 2023TwinLiteNet: An Efficient and Lightweight Model for Driveable Area and Lane Segmentation in Self-Driving CarsPaper/Code
Q-YOLOP2DICMEW 2023Q-YOLOP: Quantization-aware You Only Look Once for Panoptic Driving PerceptionPaper/Code
A-YOLOM2DTVT 2024You Only Look at Once for Real-time and Generic Multi-TaskPaper/Code
TwinLiteNetPlus2DArxiv 2024TwinLiteNetPlus: A Stronger Model for Real-time Drivable Area and Lane SegmentationPaper/Code
PETRv23DICCV 2023PETRv2: A Unified Framework for 3D Perception from Multi-Camera ImagesPaper/Code
RFTR3DECCV 2024RepVF: A Unified Vector Fields Representation for Multi-task 3D PerceptionPaper/Code

Video Lane Detection

Methods2D/3D LaneVenueTitlePaper/Code
Zou et al.2DTVT 2020Robust Lane Detection from Continuous Driving Scenes Using Deep Neural NetworksPaper/Code
Zhang et al.2DTITS 2021Lane Detection Model Based on Spatio-Temporal Network With Double Convolutional Gated Recurrent UnitsPaper/Code
MMA-Net2DICCV 2021VIL-100: A New Dataset and A Baseline Model for Video Instance Lane DetectionPaper/Code
Tabelini et al.2DIJCNN 2022Lane marking detection and classification using spatial-temporal feature poolingPaper/Code
TGC-Net2DACM MM 2022Video instance lane detection via deep temporal and geometry consistency constraintsPaper/Code
RVLD2DICCV 2023Recursive Video Lane DetectionPaper/Code
OMR2DECCV 2024OMR: Occlusion-Aware Memory-Based Refinement for Video Lane DetectionPaper/Code
ST3DLane3DBMVC 2022Spatio-Temporal Fusion-based Monocular 3D Lane DetectionPaper/Code
CurveFormre++3DArxiv 2024CurveFormer++: 3D Lane Detection by Curve Propagation with Temporal Curve Queries and AttentionPaper/Code

Online HD Map Construction

MethodsVenueTitleCode
HDMapNetICRA 2022HDMapNet: An Online HD Map Construction and Evaluation FrameworkPaper/Code
SuperFusionArxiv 2022SuperFusion: Multilevel LiDAR-Camera Fusion for Long-Range HD Map GenerationPaper/Code
VectorMapNetICML 2023VectorMapNet: End-to-end Vectorized HD Map LearningPaper/Code
MapTRICLR 2023MapTR: Structured Modeling and Learning for Online Vectorized HD Map ConstructionPaper/Code
InstaGraMCVPRW 2023InstaGraM: Instance-level Graph Modeling for Vectorized HD Map LearningPaper/Code
MachMapCVPRW 2023MachMap: End-to-End Vectorized Solution for Compact HD-Map ConstructionPaper/Code
NMPCVPR 2023Neural Map Prior for Autonomous DrivingPaper/Code
BeMapNetCVPR 2023End-to-End Vectorized HD-map Construction with Piecewise Bezier CurvePaper/Code
PivotNetICCV 2023PivotNet: Vectorized Pivot Learning for End-to-end HD Map ConstructionPaper/Code
MV-MapICCV 2023MV-Map: Offboard HD-Map Generation with Multi-view ConsistencyPaper/Code
MapSegArxiv 2023MapSeg: Segmentation guided structured model for online HD map constructionPaper/Code
NeMOArxiv 2023NeMO: Neural Map Growing System for Spatiotemporal Fusion in Bird's-Eye-View and BDD-Map BenchmarkPaper/Code
PolyDiffuseNeuIPS 2023PolyDiffuse: Polygonal Shape Reconstruction via Guided Set Diffusion ModelsPaper/Code
MapVRNeuIPS 2023Online Map Vectorization for Autonomous Driving: A Rasterization PerspectivePaper/Code
InsMapperArxiv 2023InsightMapper: A Closer Look at Inner-instance Information for Vectorized High-Definition MappingPaper/Code
MapEXArxiv 2023Mind the map! Accounting for existing map information when estimating online HDMaps from sensorPaper/Code
ScalableMapCoRL 2023ScalableMap: Scalable Map Learning for Online Long-Range Vectorized HD Map ConstructionPaper/Code
StreamMapNetWACV 2024Streammapnet: Streaming mapping network for vectorized online hd map constructionPaper/Code
MapNeXtArxiv 2024MapNeXt: Revisiting Training and Scaling Practices for Online Vectorized HD Map ConstructionPaper/Code
SQD-MapNetArxiv 2024Stream Query Denoising for Vectorized HD Map ConstructionPaper/Code
EAN-MapNetArxiv 2024EAN-MapNet: Efficient Vectorized HD Map Construction with Anchor NeighborhoodsPaper/Code
SatforHDMapICRA 2024Complementing Onboard Sensors with Satellite Map: A New Perspective for HD Map ConstructionPaper/Code
HIMapCVPR 2024HIMap: HybrId Representation Learning for End-to-end Vectorized HD Map ConstructionPaper/Code
MGMapCVPR 2024MGMap: Mask-Guided Learning for Online Vectorized HD Map ConstructionPaper/Code
HybriMapArxiv 2024HybriMap: Hybrid Clues Utilization for Effective Vectorized HD Map ConstructionPaper/Code
Shi et al.ICASSP 2024Buffered Gaussian Modeling for Vectorized HD Map ConstructionPaper/Code
GNMapSpatialDI 2024Neural HD Map Generation from Multiple Vectorized Tiles Locally Produced by Autonomous VehiclesPaper/Code
DiffMapArxiv 2024DiffMap: Enhancing Map Segmentation with Map Prior Using Diffusion ModelPaper/Code
GeMapECCV 2024Online Vectorized HD Map Construction using GeometryPaper/Code
MapQRECCV 2024Leveraging Enhanced Queries of Point Sets for Vectorized Map ConstructionPaper/Code
MapTrackerECCV 2024MapTracker: Tracking with Strided Memory Fusion for Consistent Vector HD MappingPaper/Code
ADMapECCV 2024ADMap: Anti-disturbance framework for reconstructing online vectorized HD mapPaper/Code
Mask2MapECCV 2024Mask2Map: Vectorized HD Map Construction Using Bird's Eye View Segmentation MasksPaper/Code
DTCLMapperTITS 2024DTCLMapper: Dual Temporal Consistent Learning for Vectorized HD Map ConstructionPaper/Code
MapTRv2IJCV 2024MapTRv2: An End-to-End Framework for Online Vectorized HD Map ConstructionPaper/Code
P-MapNetRAL 2024P-MapNet: Far-seeing Map Generator Enhanced by both SDMap and HDMap PriorsPaper/Code

Lane Topology Reasoning

MethodsVenueTitlePaper/Code
STSUICCV 2021Structured Bird's-Eye-View Traffic Scene Understanding from Onboard ImagesPaper/Code
TopoRoadCVPR 2022Topology Preserving Local Road Network Estimation from Single Onboard Camera ImagePaper/Code
CenterLineDetICRA 2023CenterLineDet: CenterLine Graph Detection for Road Lanes with Vehicle-mounted Sensors by Transformer for HD Map GenerationPaper/Code
TopoNetArxiv 2023Graph-based Topology Reasoning for Driving ScenesPaper/Code
Can et al.ICCV 2023Improving Online Lane Graph Extraction by Object-Lane ClusteringPaper/Code
TopoMLPICLR 2024TopoMLP: A Simple yet Strong Pipeline for Driving Topology ReasoningPaper/Code
LaneSegNetICLR 2024LaneSegNet: Map Learning with Lane Segment Perception for Autonomous DrivingPaper/Code
SMERFICRA 2024Augmenting Lane Perception and Topology Understanding with Standard Definition Navigation MapsPaper/Code
LaneGAPECCV 2024Lane Graph as Path: Continuity-preserving Path-wise Modeling for Online Lane Graph ConstructionPaper/Code
TopoLogicNeurIPS 2024TopoLogic: An Interpretable Pipeline for Lane Topology Reasoning on Driving ScenesPaper/Code

Future Direction

MethodsVenueTitlePaper/CodeDescription
Lane2SeqCVPR 2024Lane2Seq: Towards Unified Lane Detection via Sequence GenerationPaper/CodeNew lane modeling method, unified 2D lane detection; Reinforcement learning for lane detection.
M^2-3DLaneNetArxiv 2022M$^{2}$-3DLaneNet: Exploring Multi-Modal 3D Lane DetectionPaper/CodeLidar&Camera Fusion for 3D lane detection.
DV-3DLaneICLR 2024DV-3DLane: End-to-end Multi-modal 3D Lane Detection with Dual-view RepresentationPaper/CodeLidar&Camera Fusion for 3D lane detection.
WS-3D-LaneICRA 2023WS-3D-Lane: Weakly Supervised 3D Lane Detection With 2D Lane LabelsPaper/CodeWeak-supervised 3D lane detection.
MLDACVPRW 2022Multi-level Domain Adaptation for Lane DetectionPaper/CodeUnsupervised 2D lane detection
CLLDArxiv 2023Contrastive Learning for Lane Detection via cross-similarityPaper/CodeSelf-supervised 2D lane detection
Li et al.TITS 2023Robust Lane Detection through Self Pre-training with Masked Sequential Autoencoders and Fine-tuning with Customized PolyLossPaper/CodeSelf-supervised 2D lane detection
LaneCorrectArxiv 2024LaneCorrect: Self-supervised Lane DetectionPaper/CodeSelf-supervised 2D lane detection