Research map / Detection and segmentation / CNN features
Rethinking Efficient Lane Detection via Curve Modeling
Zhengyang Feng, Shaohua Guo, Xin Tan, Ke Xu, Min Wang, Lizhuang Ma
This paper tried a common change: Add data augmentation
Add strong data augmentation · CULane test · F1: 55.11 -> 75.41 Improved
What 689 accepted papers report for this change
- 560 improved
- 44 no clear effect
- 26 mixed
- 59 hurt
- Understanding Multimodal Learning: A Loss Landscape Smoothness Perspective ICML 2026 · add data augmentation (jitter, crop, flip) · UPMC-Food101 · accuracy: 92.84 -> 94.93Improved
- AsyncBEV: Cross-modal flow alignment in Asynchronous 3D Object Detection ICLR 2026 · add EMC + asynchronous data augmentation training · nuScenes val, 0 s offset · NDS: 66.7 -> 63.1Hurt
- Towards Understanding Continual Factual Knowledge Acquisition of Language Models: From Theory to Algorithm ICML 2026 · add data augmentation (5 biographies per individual) in pre-training · Biography PT test, unseen templates · EM: 2.71 -> 81.27Improved
See this change in detailMatched by the wording of each paper's own result, read by AI from the full text. Papers report gains more often than failures, so this shows what gets published, not the chance a change works.
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use on-image cubic Bézier curves instead of 3rd-order polynomial curves (ablation) | BézierLaneNet · CULane test | F1: 5.01 -> 75.41 | Improved |
| Use sampling loss instead of direct control-point loss (ablation) | Bézier curve fitting · CULane test | F1: 63.74 -> 68.89 | Improved |
| Add feature flip fusion (ablation) | BézierLaneNet · CULane test | F1: 68.89 -> 70.28 | Improved |
| Add auxiliary segmentation loss without feature flip fusion (ablation) | Sampling-loss baseline · CULane test | F1: 68.89 -> 65.82 | Hurt |
| Add auxiliary segmentation loss alongside feature flip fusion (ablation) | BézierLaneNet · CULane test | F1: 70.28 -> 73.97 | Improved |
| Use deformable convolution to align flipped features (ablation) | Feature flip fusion · CULane test | F1: 70.28 -> 72.96 | Improved |
| Use deformable convolution rather than standard convolution in feature flip fusion with auxiliary segmentation (ablation) | BézierLaneNet · CULane test | F1: 73.97 -> 75.41 | Improved |
| Add strong data augmentation (ablation) | BézierLaneNet (ResNet-34) · CULane test | F1: 55.11 -> 75.41 | Improved |
| Add strong data augmentation (ablation) | LSTR (ResNet-18, 2×) · CULane test | F1: 39.77 -> 68.72 | Improved |
| Add feature flip fusion (ablation) | Bézier baseline · TuSimple test | Accuracy: 93.36 -> 95.26 (+ 1.90) | Improved |
| Add feature flip fusion (ablation) | Bézier baseline · LLAMAS val | F1: 95.27 -> 96.00~(+ 0.73) | Improved |
| Apply strong augmentation to segmentation-based methods (ablation) | Segmentation-based lane detectors · TuSimple | performance: 1~2% | Hurt |
Closest papers by meaning
- Bezier Everywhere All at Once: Learning Drivable Lanes as Bezier Graphs (CVPR 2024 · similarity 0.89)
- Sketch and Refine: Towards Fast and Accurate Lane Detection (AAAI 2024 · similarity 0.89)
- CurveLane-NAS: Unifying Lane-Sensitive Architecture Search and Adaptive Point Blending (ECCV 2020 · similarity 0.89)
- LaneCPP: Continuous 3D Lane Detection using Physical Priors (CVPR 2024 · similarity 0.88)
- Rethinking Lanes and Points in Complex Scenarios for Monocular 3D Lane Detection (CVPR 2025 · similarity 0.88)
- A Siamese Transformer with Hierarchical Refinement for Lane Detection (NeurIPS 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
