Research map / Detection and segmentation / CNN features
Lane detection
Detect lanes with transformers, global-local perception, and sequence generation.
32 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · an approach in CNN features, keyword cluster “local · context · semantic segmentation”
State History Benchmarks Tried Checklist Preprints Who Export
Where it stands
Quiet None of the tracked benchmarks has a report from 2025–26.
11 papers in 2023–24 and 4 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 3.4% to 1.2%.
No longer reported by papers here: CULane test (F1, last in 2024); TuSimple test (Accuracy, last in 2024); TuSimple test (F1 (%), last in 2024).
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2017 · MCMLSD: A Dynamic Programming Approach to Line Segment Detection (CVPR 2017 · 92 citations)
- 2018 · RoadTracer: Automatic Extraction of Road Networks From Aerial Images (CVPR 2018 · 373 citations)
- 2019 · Improved Road Connectivity by Joint Learning of Orientation and Segmentation (CVPR 2019 · 240 citations)
- 2020 · "Looking at the Right Stuff" - Guided Semantic-Gaze for Autonomous Driving (CVPR 2020 · 51 citations)
- 2021 · Keep Your Eyes on the Lane: Real-Time Attention-Guided Lane Detection (CVPR 2021 · 472 citations)
- 2022 · CLRNet: Cross Layer Refinement Network for Lane Detection (CVPR 2022 · 313 citations)
- 2023 · ReasonNet: End-to-End Driving With Temporal and Global Reasoning (CVPR 2023 · 105 citations)
- 2024 · Lane2Seq: Towards Unified Lane Detection via Sequence Generation (CVPR 2024 · 41 citations)
- 2025 · Improving Transformer Based Line Segment Detection with Matched Predicting and Re-ranking (AAAI 2025 · 4 citations)
- 2026 · Online Navigation Refinement: Achieving Lane-Level Guidance by Associating Standard-Definition and Online Perception Maps (ICLR 2026 · 0 citations)
Earliest papers here
- MCMLSD: A Dynamic Programming Approach to Line Segment Detection (CVPR 2017 · 92 citations)
- RoadTracer: Automatic Extraction of Road Networks From Aerial Images (CVPR 2018 · 373 citations)
- Learning Integrated Holism-Landmark Representations for Long-Term Loop Closure Detection (AAAI 2018 · 12 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
CULane test · F1 (higher is better)
5 papers, 2022–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 77.06 | Laneformer: Object-Aware Row-Column Transformers for Lane Detection (AAAI 2022 · 71 citations) |
| 2024 | 81.83 | A Hybrid Global-Local Perception Network for Lane Detection (AAAI 2024 · 12 citations) |
TuSimple test · Accuracy (higher is better)
4 papers, 2022–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 96.8 | Laneformer: Object-Aware Row-Column Transformers for Lane Detection (AAAI 2022 · 71 citations) |
| 2024 | 96.89 | A Hybrid Global-Local Perception Network for Lane Detection (AAAI 2024 · 12 citations) |
TuSimple test · F1 (%) (higher is better)
4 papers, 2022–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 97.89 | CLRNet: Cross Layer Refinement Network for Lane Detection (CVPR 2022 · 313 citations) |
| 2024 | 97.98 | GSENet:Global Semantic Enhancement Network for Lane Detection (AAAI 2024 · 23 citations) |
What has been tried
30 papers here report 324 measured changes (a component added, removed or swapped, compared with a baseline): helped 74%, no clear effect 4%, hurt 20%, mixed 2%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| add MFRL to segmentation-format Lane2Seq | CULane · F1 (%) | 64.27 -> 79.64 (+15.37) | Helped | Lane2Seq: Towards Unified Lane Detection via Sequence Generation (CVPR 2024 · 41 citations) |
| Add edge-granularity embedding in the frequency domain | BSDS500 · ODS | 0.829 -> 0.846 | Helped | MuGE: Multiple Granularity Edge Detection (CVPR 2024 · 38 citations) |
| use direction-based proposal initialization instead of anchor-based initialization | CurveLanes subset · detection performance | Helped | Sketch and Refine: Towards Fast and Accurate Lane Detection (AAAI 2024 · 27 citations) | |
| add Angle Loss | CULane · F1@50 | 79.58 -> 79.81 | Helped | GSENet:Global Semantic Enhancement Network for Lane Detection (AAAI 2024 · 23 citations) |
| add MFA | CULane · mF1 | 52.84% to 54.58% | Helped | A Hybrid Global-Local Perception Network for Lane Detection (AAAI 2024 · 12 citations) |
| use lane segment learning instead of the multi-branch CL + ME formulation | OpenLane-V2 centerline perception · OLS | +4.8 on OLS | Helped | LaneSegNet: Map Learning with Lane Segment Perception for Autonomous Driving (ICLR 2024 · 10 citations) |
| Add Semantic Guided Focus | CULane, W4A4 · F1 score | 73.27 -> 76.31 | Helped | Selective Focus: Investigating Semantics Sensitivity in Post-training Quantization for Lane Detection (AAAI 2024 · 5 citations) |
| Add matched predicting | Wireframe · sAP5 | 50.3 -> 60.2 | Helped | Improving Transformer Based Line Segment Detection with Matched Predicting and Re-ranking (AAAI 2025 · 4 citations) |
| add GFEM | CULane · F1@50 | 81.64 -> 81.81 | Helped | GEONet: Global Enhancement and Optimization Network for Lane Detection (AAAI 2025 · 2 citations) |
| add the full multi-fold top-down guidance (TRG and DISG) to bottom-up features | TOI · AP | 20 -> 60 | Helped | On-Road Object Importance Estimation: A New Dataset and A Model with Multi-Fold Top-Down Guidance (NeurIPS 2024 · 1 citation) |
| Add high-to-low refinement from p0 to p3 instead of refining at p0 only | CULane · F1 score | 76.87 -> 80.01 | Helped | A Siamese Transformer with Hierarchical Refinement for Lane Detection (NeurIPS 2024) |
What reviewers will expect
Datasets and baselines named in the 12 papers of this approach from 2024–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- CULane · 7 of 12 papers
- TuSimple · 7 of 12 papers
- LLAMAS · 2 of 12 papers
- CurveLanes · 2 of 12 papers
Baselines
- LaneATT · 6 of 12 papers
- CLRNet · 6 of 12 papers
- RESA · 5 of 12 papers
- SCNN · 4 of 12 papers
- CondLaneNet · 4 of 12 papers
- UFLD · 3 of 12 papers
- LaneAF · 3 of 12 papers
- UFLDv2 · 3 of 12 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 15 papers; from OpenAlex and the papers' affiliations).
- Fudan University · 2 papers
- Southeast University · 2 papers
- Tsinghua University · 2 papers
- SenseTime · 2 papers
- Microsoft · 2 papers
Use it
Download BibTeX of 29 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Context modeling for semantic segmentation (241 papers)
- Local features and context in CNNs (170 papers)
- Context for hard detection conditions (133 papers)
- Domain-specific segmentation and detection (118 papers)
- Forgery detection (46 papers)
Elsewhere in the topic
- Region-based and query-based detectors (183 papers)
- Convolution-transformer backbones (101 papers)
- Box regression and loss design (94 papers)
- Efficient DETR query design (85 papers)
- Boundary-aware instance segmentation (83 papers)
- Plain and high-resolution ViTs (69 papers)
- Arbitrary-shape scene text detection (61 papers)
- RGB salient object detection (61 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).