Ultra-faster Line Detection for lane detection and belt line detection. The fastest lightweight lane line detection algorithm! Work step by step open.
Please see INSTALL.md
python xxxx.py configs/xxx.pydeviation.py : deviation for beltline， test.py : test accuarcy and speed. Code for training will be released in future.
We provide trained models on Tusimple at r=0.6. Other model will be released in future.
| Dataset | Metric This repo | Avg FPS on GTX 2080Ti | Avg FPS on NJNX | Model |
|---|---|---|---|---|
| Tusimple | 95.62 | 682 | 115 | GoogleDrive |
Thanks (UFAST) https://github.com/cfzd/Ultra-Fast-Lane-Detection and (RepVGG) https://github.com/DingXiaoH/RepVGG. Please use following citation to show respect.
@InProceedings{qin2020ultra,
author = {Qin, Zequn and Wang, Huanyu and Li, Xi},
title = {Ultra Fast Structure-aware Deep Lane Detection},
booktitle = {The European Conference on Computer Vision (ECCV)},
year = {2020}
}
@inproceedings{ding2021repvgg,
title={Repvgg: Making vgg-style convnets great again},
author={Ding, Xiaohan and Zhang, Xiangyu and Ma, Ningning and Han, Jungong and Ding, Guiguang and Sun, Jian},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
pages={13733--13742},
year={2021}
}If you use the techniques in this article please cite:
@article{chen2023non,
title={Non-Residual Unrestricted Pruned Ultra-faster Line Detection for Edge Devices},
author={Chen, Pengpeng and Liu, Dongjingdian and Gao, Shouwan},
journal={Pattern Recognition},
pages={109321},
year={2023},
publisher={Elsevier}
}