Skip to content

Commit 6999234

Browse files
hctian713hctian713
committed
[release] data and checkpoints
1 parent cea2930 commit 6999234

File tree

2 files changed

+199
-39
lines changed

2 files changed

+199
-39
lines changed

‎README.md‎

Lines changed: 148 additions & 39 deletions
Original file line numberDiff line numberDiff line change
@@ -7,63 +7,187 @@
77

88
# **Learning to Drive via Real-World Simulation at Scale**
99

10-
[![Paper](https://img.shields.io/badge/ArXiv-2511.23369-A42C25?style=for-the-badge&logo=arxiv&logoColor=white)](https://arxiv.org/abs/2511.23369)
11-
[![Home](https://img.shields.io/badge/page-project-5F259F?style=for-the-badge&logo=homepage&logoColor=white)](https://opendrivelab.com/SimScale/)
12-
[![Hugging Face](https://img.shields.io/badge/PAGE-HF-fcd022?style=for-the-badge&logo=huggingface&logoColor=white)](https://huggingface.co/papers/2511.23369)
13-
[![License](https://img.shields.io/badge/License-Apache--2.0-2380C1?style=for-the-badge&)](https://github.com/OpenDriveLab/SimScale/blob/main/LICENSE)
10+
[![Paper](https://img.shields.io/badge/ArXiv-A42C25?style=for-the-badge&logo=arxiv&logoColor=white)](https://arxiv.org/abs/2511.23369)
11+
[![Home](https://img.shields.io/badge/project_page-5F259F?style=for-the-badge&logo=homepage&logoColor=white)](https://opendrivelab.com/SimScale/)
12+
[![Hugging Face](https://img.shields.io/badge/hugging_face-ffc107?style=for-the-badge&logo=huggingface&logoColor=white)](https://huggingface.co/datasets/OpenDriveLab/SimScale)
13+
[![License](https://img.shields.io/badge/Apache--2.0-2380C1?style=for-the-badge&)](https://github.com/OpenDriveLab/SimScale/blob/main/LICENSE)
1414

1515

16+
</div>
17+
1618

19+
<div id="top" align="center">
20+
<p align="center">
21+
<img src="assets/teaser.png" >
22+
</p>
1723
</div>
1824

1925

26+
2027
> [Haochen Tian](https://github.com/hctian713),
2128
> [Tianyu Li](https://github.com/sephyli),
2229
> [Haochen Liu](https://georgeliu233.github.io/),
2330
> [Jiazhi Yang](https://github.com/YTEP-ZHI),
2431
> [Yihang Qiu](https://github.com/gihharwtw),
25-
> Guang Li,
26-
> Junli Wang,
32+
> [Guang Li](https://scholar.google.com/citations?user=McEfO8UAAAAJ&hl=en),
33+
> [Junli Wang](https://openreview.net/profile?id=%7EJunli_Wang4),
2734
> [Yinfeng Gao](https://scholar.google.com/citations?user=VTn0hqIAAAAJ&hl=en),
2835
> [Zhang Zhang](https://scholar.google.com/citations?user=rnRNwEMAAAAJ&hl=en),
2936
> [Liang Wang](https://scholar.google.com/citations?user=8kzzUboAAAAJ&hl=en),
30-
> Hangjun Ye,
37+
> [Hangjun Ye](https://scholar.google.com/citations?user=68tXhe8AAAAJ&hl=en),
3138
> [Tieniu Tan](https://scholar.google.com/citations?user=W-FGd_UAAAAJ&hl=en),
3239
> [Long Chen](https://long.ooo/),
3340
> [Hongyang Li](https://lihongyang.info/)
3441
>
3542
>
36-
> 📧 Primary Contact: Haochen Tian (tianhaochen2023@ia.ac.cn)
37-
<!-- >
38-
> 📜 [[technical report](https://arxiv.org/abs/2511.23369)],
39-
> 🏠 [[project page](https://opendrivelab.com/SimScale)] -->
40-
41-
<div id="top" align="center">
42-
<p align="center">
43-
<img src="assets/teaser.png" >
44-
</p>
45-
</div>
43+
> - 📧 Primary Contact: Haochen Tian (tianhaochen2023@ia.ac.cn)
44+
> - 📜 Materials: 🌐 [𝕏](https://x.com/OpenDriveLab/status/1999507869633527845) | 📰 [Media](https://mp.weixin.qq.com/s/OGV3Xlb0bHSSSloG11qFJA) | 🗂️ [Slides TODO]()
4645
47-
## 🔥 Highlights
46+
---
4847

48+
## 🔥 Highlights
4949

50-
- 🏗️ A scalable simulation pipeline that synthesizes diverse and high-fidelity reactive driving scenarios with pseudo-expert demonstrations.
50+
- 🏗️ A scalable simulation pipepline that synthesizes diverse and high-fidelity reactive driving scenarios with pseudo-expert demonstrations.
5151
- 🚀 An effective sim-real co-training strategy that improves robustness and generalization synergistically across various end-to-end planners.
5252
- 🔬 A comprehensive recipe that reveals crucial insights into the underlying scaling properties of sim-real learning systems for end-to-end autonomy.
5353

5454

5555
## 📢 News
5656

57-
- **[2025/12/1]** We released our [paper](https://arxiv.org/abs/2511.23369) on arXiv.
57+
- **`[2025/12/31]`** We released the data, and models **v1.0**. Happy New Year ! 🎄
58+
- **`[2025/12/1]`** We released our [paper](https://arxiv.org/abs/2511.23369) on arXiv.
59+
5860

5961
## 📋 TODO List
6062

61-
- [ ] Simulation Data release (Dec. 2025).
62-
- [ ] Sim-Real Co-training Code release (Dec. 2025).
63-
- [ ] Checkpoints release (Dec. 2025).
63+
- [ ] Sim-Real Co-training Code release (Jan. 2026).
64+
- [x] Simulation Data release (Dec. 2025).
65+
- [x] Checkpoints release (Dec. 2025).
66+
67+
---
68+
69+
## 📌 Table of Contents
70+
71+
<!-- - [📢 News](#news)
72+
- [📋 TODO List](#todo-list) -->
73+
- 🤗 [Model Zoo](#-model-zoo)
74+
<!-- - 🎯 [Getting Started](#-getting-started) -->
75+
- 📦 [Data Preparation](#-data-preparation)
76+
- [Download Dataset](#1-download-dataset)
77+
<!-- - [Set Up Configuration](#2-set-up-configuration) -->
78+
<!-- - ⚙️ [Sim-Real Co-Training](#-sim-real-co-training-recipe)
79+
- [Co-Training with Pseudo-Expert](#co-training-with-pseudo-expert)
80+
- [Co-Training with Rewards Only](#co-training-with-rewards-only)
81+
- 🔍 [Inference](#-inference)
82+
- [NAVSIM v2 navhard](#navsim-v2-navhard)
83+
- [NAVSIM v2 navtest](#navsim-v2-navtest) -->
84+
- ⭐ [License and Citation](#-license-and-citation)
85+
86+
## 🤗 Model Zoo
87+
88+
<table>
89+
<tr style="text-align: center;">
90+
<th rowspan="2">Model</th>
91+
<th rowspan="2">Backbone</th>
92+
<th rowspan="2">Sim-Real Config</th>
93+
<th colspan="2">NAVSIM v2 navhard</th>
94+
<th colspan="2">NAVSIM v2 navtest</t>
95+
96+
</tr>
97+
98+
<tr style="text-align: center;">
99+
<th>EPDMS</th>
100+
<th>CKPT</th>
101+
<th>EPDMS</th>
102+
<th>CKPT</th>
103+
</tr>
104+
105+
<!-- LTF -->
106+
<tr>
107+
<td><a href="#">LTF</a></td>
108+
<td>ResNet34</td>
109+
<td>w/ pseudo-expert</td>
110+
<td>30.3 | +6.9</td>
111+
<td><a href="https://huggingface.co/datasets/OpenDriveLab/SimScale/blob/main/SimScale_ckpts/LTF/ltf_sim_navhard.ckpt">Link</a></td>
112+
<td>84.4 | +2.9</td>
113+
<td><a href="https://huggingface.co/datasets/OpenDriveLab/SimScale/blob/main/SimScale_ckpts/LTF/ltf_sim_navtest.ckpt">Link</a></td>
114+
</tr>
115+
116+
<!-- DiffusionDrive -->
117+
<tr>
118+
<td><a href="#">DiffusionDrive</a></td>
119+
<td>ResNet34</td>
120+
<td>w/ pseudo-expert</td>
121+
<td>32.6 | +5.1</td>
122+
<td><a href="https://huggingface.co/datasets/OpenDriveLab/SimScale/blob/main/SimScale_ckpts/DiffusionDrive/diffusiondrive_sim_navhard.ckpt">Link</a></td>
123+
<td>85.9 | +1.7</td>
124+
<td><a href="https://huggingface.co/datasets/OpenDriveLab/SimScale/blob/main/SimScale_ckpts/DiffusionDrive/diffusiondrive_sim_navtest.ckpt">Link</a></td>
125+
</tr>
126+
127+
<!-- GTRS-Dense block -->
128+
<tr>
129+
<td rowspan="4"><a href="#">GTRS-Dense</a></td>
130+
<td rowspan="2">ResNet34</td>
131+
<td>w/ pseudo-expert</td>
132+
<td>46.1 | +7.8</td>
133+
<td><a href="https://huggingface.co/datasets/OpenDriveLab/SimScale/blob/main/SimScale_ckpts/GTRS_Dense/gtrs_dense_resnet_sim_expert_navhard.ckpt">Link</a></td>
134+
<td>84.0 | +1.7</td>
135+
<td><a href="https://huggingface.co/datasets/OpenDriveLab/SimScale/blob/main/SimScale_ckpts/GTRS_Dense/gtrs_dense_resnet_sim_expert_navtest.ckpt">Link</a></td>
136+
</tr>
137+
138+
<tr>
139+
<td>rewards only</td>
140+
<td>46.9 | +8.6</td>
141+
<td><a href="https://huggingface.co/datasets/OpenDriveLab/SimScale/blob/main/SimScale_ckpts/GTRS_Dense/gtrs_dense_resnet_sim_reward_navhard.ckpt">Link</a></td>
142+
<td>84.6 | +2.3</td>
143+
<td><a href="https://huggingface.co/datasets/OpenDriveLab/SimScale/blob/main/SimScale_ckpts/GTRS_Dense/gtrs_dense_resnet_sim_reward_navtest.ckpt">Link</a></td>
144+
</tr>
145+
146+
<tr>
147+
<td rowspan="2">V2-99</td>
148+
<td>w/ pseudo-expert</td>
149+
<td>47.7 | +5.8</td>
150+
<td><a href="https://huggingface.co/datasets/OpenDriveLab/SimScale/blob/main/SimScale_ckpts/GTRS_Dense/gtrs_dense_vov_sim_expert_navhard.ckpt">Link</a></td>
151+
<td>84.5 | +0.5</td>
152+
<td><a href="https://huggingface.co/datasets/OpenDriveLab/SimScale/blob/main/SimScale_ckpts/GTRS_Dense/gtrs_dense_vov_sim_expert_navtest.ckpt">Link</a></td>
153+
</tr>
154+
155+
<tr>
156+
<td>rewards only</td>
157+
<td>48.0 | +6.1</td>
158+
<td><a href="https://huggingface.co/datasets/OpenDriveLab/SimScale/blob/main/SimScale_ckpts/GTRS_Dense/gtrs_dense_vov_sim_reward_navhard.ckpt">Link</a></td>
159+
<td>84.8 | +0.8</td>
160+
<td><a href="https://huggingface.co/datasets/OpenDriveLab/SimScale/blob/main/SimScale_ckpts/GTRS_Dense/gtrs_dense_vov_sim_reward_navtest.ckpt">Link</a></td>
161+
</tr>
162+
</table>
163+
164+
> [!NOTE]
165+
> We fixed a minor error in the simulation process without changing the method, resulting in better performance than the numbers reported in the arXiv version. We will update the arXiv paper soon.
166+
167+
168+
169+
## 📦 Data Preparation
170+
171+
Our released simulation data is based on [nuPlan](https://www.nuscenes.org/nuplan) and [NAVSIM](https://github.com/autonomousvision/navsim). We recommend first preparing the real-world data by following the instructions in [Download NAVSIM](https://github.com/autonomousvision/navsim/blob/main/docs/install.md#2-download-the-dataset).
172+
173+
### 1. Download Dataset
174+
175+
Our simulation data format follows that of [OpenScene](https://github.com/OpenDriveLab/OpenScene/blob/main/docs/getting_started.md#download-data), with each clip/log has a fixed temporal horizon of 6 seconds (2s history + 4s future).
176+
177+
We provide [scripts](./docs/download.sh) for downloading the simulation data.
64178

65179

66-
## ⭐ Citation
180+
181+
## ❤️ Acknowledgements
182+
183+
We acknowledge all the open-source contributors for the following projects to make this work possible:
184+
185+
- [NAVSIM](https://github.com/autonomousvision/navsim) | [MTGS](https://github.com/OpenDriveLab/MTGS) | [GTRS](https://github.com/NVlabs/GTRS) | [DiffusionDrive](https://github.com/hustvl/DiffusionDrive)
186+
187+
## ⭐ License and Citation
188+
189+
All content in this repository is under the [Apache-2.0 license](https://www.apache.org/licenses/LICENSE-2.0).
190+
The released data is based on [nuPlan](https://www.nuscenes.org/nuplan) and is under the [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/) license.
67191

68192
If any parts of our paper and code help your research, please consider citing us and giving a star to our repository.
69193

@@ -75,20 +199,5 @@ If any parts of our paper and code help your research, please consider citing us
75199
year={2025}
76200
}
77201
```
78-
## ⚖️ License
79-
80-
All content in this repository is under the [Apache-2.0 license](https://www.apache.org/licenses/LICENSE-2.0).
81-
The released data is based on [nuPlan](https://www.nuscenes.org/nuplan) and are under the [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/) license.
82-
83-
## ❤️ Acknowledgements
84-
85-
We acknowledge all the open-source contributors for the following projects to make this work possible:
86-
87-
- [NAVSIM](https://github.com/autonomousvision/navsim)
88-
- [MTGS](https://github.com/OpenDriveLab/MTGS)
89-
- [GTRS](https://github.com/NVlabs/GTRS)
90-
- [DiffusionDrive](https://github.com/hustvl/DiffusionDrive)
91-
92-
93202

94203

‎docs/download.sh‎

Lines changed: 51 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,51 @@
1+
repo="https://huggingface.co/datasets/OpenDriveLab/SimScale/resolve/main"
2+
3+
# simulation data with planner-based pseudo-expert
4+
rounds=5
5+
splits=(66,56,47,39,33)
6+
7+
for round in $(seq 0 $((rounds - 1))); do
8+
# 1. meta data
9+
wget ${repo}/SimScale_data/synthetic_reaction_pdm_v1.0-${round}/simscale_pdm_v1.0-${round}_meta_datas.tar.gz
10+
echo "[Planner-based] Downloaded meta data for round ${round}"
11+
tar -xzvf simscale_pdm_v1.0-${round}_meta_datas.tar.gz
12+
rm simscale_pdm_v1.0-${round}_meta_datas.tar.gz
13+
14+
# 2. hist sensor data
15+
split=${splits[$round]}
16+
for idx in $(seq 0 $((split - 1))); do
17+
wget ${repo}/SimScale_data/synthetic_reaction_pdm_v1.0-${round}/simscale_pdm_v1.0-${round}_sensor_blobs_hist/simscale_pdm_v1.0-${round}_sensor_blobs_hist_${idx}.tar.gz
18+
echo "[Planner-based] Downloaded hist sensor data ${idx} for round ${round}"
19+
tar -xzvf simscale_pdm_v1.0-${round}_sensor_blobs_hist_${idx}.tar.gz
20+
rm simscale_pdm_v1.0-${round}_sensor_blobs_hist_${idx}.tar.gz
21+
done
22+
23+
# 3. future sensor data
24+
# TBD
25+
26+
done
27+
28+
# simulation data with recovery-based pseudo-expert
29+
rounds=5
30+
splits=(45,36,28,22,17)
31+
32+
for round in $(seq 0 $((rounds - 1))); do
33+
# 1. meta data
34+
wget ${repo}/SimScale_data/synthetic_reaction_recovery_v1.0-${round}/simscale_recovery_v1.0-${round}_meta_datas.tar.gz
35+
echo "[Recovery-based] Downloaded meta data for round ${round}"
36+
tar -xzvf simscale_recovery_v1.0-${round}_meta_datas.tar.gz
37+
rm simscale_recovery_v1.0-${round}_meta_datas.tar.gz
38+
39+
# 2. hist sensor data
40+
split=${splits[$round]}
41+
for idx in $(seq 0 $((split - 1))); do
42+
wget ${repo}/SimScale_data/synthetic_reaction_recovery_v1.0-${round}/simscale_recovery_v1.0-${round}_sensor_blobs_hist/simscale_recovery_v1.0-${round}_sensor_blobs_hist_${idx}.tar.gz
43+
echo "[Recovery-based] Downloaded hist sensor data ${idx} for round ${round}"
44+
tar -xzvf simscale_recovery_v1.0-${round}_sensor_blobs_hist_${idx}.tar.gz
45+
rm simscale_recovery_v1.0-${round}_sensor_blobs_hist_${idx}.tar.gz
46+
done
47+
48+
# 3. future sensor data
49+
# TBD
50+
51+
done

0 commit comments

Comments
 (0)