ReconDreamer: Crafting World Models for Driving Scene Reconstruction via Online Restoration

Project Page | Paper

News

• [2024/12/04] 🚀 We have released the weights for some scenarios based on Street Gaussians, as well as the NTA-IoU calculation script.
• [2024/11/29] 🚀 Repository Initialization.

Abstract

Closed-loop simulation is crucial for end-to-end autonomous driving. Existing sensor simulation methods (e.g., NeRF and 3DGS) reconstruct driving scenes based on conditions that closely mirror training data distributions. However, these methods struggle with rendering novel trajectories, such as lane changes. Recent works have demonstrated that integrating world model knowledge alleviates these issues. Despite their efficiency, these approaches still encounter difficulties in the accurate representation of more complex maneuvers, with multi-lane shifts being a notable example.Therefore, we introduce ReconDreamer, which enhances driving scene reconstruction through incremental integration of world model knowledge. Specifically, DriveRestorer is proposed to mitigate artifacts via online restoration. This is complemented by a progressive data update strategy designed to ensure high-quality rendering for more complex maneuvers. To the best of our knowledge, ReconDreamer is the first method to effectively render in large maneuvers. Experimental results demonstrate that ReconDreamer outperforms Street Gaussians in the NTA-IoU, NTL-IoU, and FID, with relative improvements by 24.87%, 6.72%, and 29.97%. Furthermore, ReconDreamer surpasses DriveDreamer4D with PVG during large maneuver rendering, as verified by a relative improvement of 195.87% in the NTA-IoU metric and a comprehensive user study.

ReconDreamer Framework

Scenario Selection

All selected scenes are sourced from the validation set of the Waymo dataset. The official file names of these scenes, are listed along with their respective starting and ending frames.

Scene	Start Frame	End Frame
segment-10359308928573410754_720_000_740_000_with_camera_labels.tfrecord	120	159
segment-11450298750351730790_1431_750_1451_750_with_camera_labels.tfrecord	0	39
segment-12496433400137459534_120_000_140_000_with_camera_labels.tfrecord	110	149
segment-15021599536622641101_556_150_576_150_with_camera_labels.tfrecord	0	39
segment-16767575238225610271_5185_000_5205_000_with_camera_labels.tfrecord	0	39
segment-17860546506509760757_6040_000_6060_000_with_camera_labels.tfrecord	90	129
segment-3015436519694987712_1300_000_1320_000_with_camera_labels.tfrecord	40	79
segment-6637600600814023975_2235_000_2255_000_with_camera_labels.tfrecord	70	109

Install

conda create -n recondreamer python=3.8
conda activate recondreamer
pip install torch==1.13.1+cu116 torchvision==0.14.1+cu116 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu116
pip install -r requirments.txt
pip install ./submodules/diff-gaussian-rasterization
pip install ./submodules/simple-knn
pip install ./submodules/simple-waymo-open-dataset-reader
python script/test_gaussian_rasterization.py

Demo

Scene	Weight	Data
segment-10359308928573410754_720_000_740_000_with_camera_labels.tfrecord	Link to Weight	Link to Data

Recondreamer/
├── config/
│   ├── street_gaussians.yaml
│   └── recondreamer.yaml
|── data
├── .....
├── output/
│    └── waymo_full_exp/005
│        ├── recondreamer
│        └── street_gaussians
│── render.py

After downloading the data and weights, place the Data contents into the data/ folder, and move the Weight files into the output/ directory.

Rendering

python render.py --config configs/recondreamer.yaml
python render.py --config configs/street_gaussians.yaml

Calculate the NTA-IoU metric

Model	Weight
Yolo11x.pt	Link to Weight

python script/NTAIou/script/GT.py
python script/NTAIou/script/detect.py
python script/NTAIou/script/calculate.py

The detailed metrics will be saved in the file: script/NTAIou/average_iou_results_005.txt.

Rendering Results in Lane Shift @ 3m Novel Trajectory

combined_164_3.mp4

combined_27_3.mp4

combined_18_3.mp4

combined_05_3.mp4

Rendering Results in Lane Shift @ 6m Novel Trajectory

combined_164_6.mp4

combined_27_6.mp4

combined_18_6.mp4

combined_05_6.mp4

Acknowledgements

We would like to thank the following works and projects, for their open research and exploration: Street Gaussians, DriveStudio, DriveDreamer, and DriveDreamer4D.

Citation

If you find this code useful for your research, please use the following BibTeX entry.

@inproceedings{Ni2024ReconDreamerCW,
    title={ReconDreamer: Crafting World Models for Driving Scene Reconstruction via Online Restoration},
    author={Chaojun Ni and Guosheng Zhao and Xiaofeng Wang and Zheng Zhu and Wenkang Qin and Guan Huang and Chen Liu and Yuyin Chen and Yida Wang and Xueyang Zhang and Yifei Zhan and Kun Zhan and Peng Jia and Xianpeng Lang and Xingang Wang and Wenjun Mei},
    year={2024},
    url={https://arxiv.org/abs/2411.19548}
}