UnLanedet: A Unified Lane Detection Toolbox
- Publicada
- Servidor
- Preprints.org
- DOI
- 10.20944/preprints202507.1610.v1
Lane detection is an important topic in autonomous driving and computer vision. However, the community currently lacks a unified and comprehensive benchmark specifically tailored for lane detection models. To address this issue, we develop a unified, highly modular, and lightweight codebase named UnLanedet, supporting a majority of the mainstream lane detection methods and datasets. We benchmark the lane detection methods under UnLanedet and provide fair comparison results. Besides, we optimize the training loop to enhance the model performance. We hope that UnLanedet can benefit the lane detection community and offer a unified platform to compare different lane detection models. Our code is available at https://github.com/zkyntu/UnLanedet.