Hardliu/nuCarla
收藏Hugging Face2025-12-14 更新2026-03-29 收录
下载链接：
https://hf-mirror.com/datasets/Hardliu/nuCarla
下载链接
链接失效反馈官方服务：
资源简介：
---
viewer: false
license: apache-2.0
size_categories:
- 100M<n<1B
---
nuCarla is a large-scale, nuScenes-compatible, camera-based perception dataset developed in the CARLA simulator, featuring 9 distinct maps, 14 weather conditions, and 6 object classes. It is designed to facilitate training of effective perception representations for end-to-end autonomous driving development. The dataset contains 1,000 scenarios in total, with 700 for training, 150 for validation, and 150 for testing, matching the nuScenes split. To thoroughly validate nuCarla, we train and evaluate four state-of-the-art Bird's Eye View (BEV) perception models: [BEVFormer](https://github.com/fundamentalvision/BEVFormer), [PETR](https://github.com/megvii-research/PETR), [BEVDet](https://github.com/HuangJunJie2017/BEVDet), [FastBEV](https://github.com/ymlab/advanced-fastbev), and demonstrate that all models achieve competitive performance using the official nuScenes detection metrics.
GitHub: https://github.com/michigan-traffic-lab/nuCarla/tree/main.
If you find this dataset helpful, please give us a star on GitHub!
提供机构：
Hardliu
