Cosmos-Transfer-LidarGen-Example

## Dataset Description:
This dataset consists of 10 synchronized clips of LiDAR and multi-view video, accompanied by corresponding camera and LiDAR poses.

This dataset is ready for commercial/non-commercial use.

## Dataset Owner(s):
NVIDIA Corporation

## Dataset Creation Date:
Oct 23, 2025

## License/Terms of Use:
This dataset is governed by [NVIDIA Autonomous Vehicle Dataset License Agreement](https://huggingface.co/datasets/nvidia/Cosmos-Transfer-LidarGen-Example/blob/main/LICENSE).

## Intended Usage:
This dataset is intended to demonstrate and facilitate understanding and usage of the Cosmos-Transfer-LidarGEN model.

## Dataset Characterization
** Data Collection Method<br>
* [Automatic/Sensors] <br>

** Labeling Method<br>
* [Automatic/Sensors] <br>

## Dataset Format
Modality: LiDAR, Camera Intrinsics and Extrinsics, multi-view videos

Format Classification: LiDAR Point Clouds, multi-view videos

You will see the following folder structure:
Sensor Data
| Folder | File Format | Description |
| :--- | :--- | :--- |
| lidar | .tar | LiDAR point clouds without motion compensation (10 FPS) |
| ftheta\_camera\_front\_wide\_120fov | .mp4 | Video captured by front wide camera |
| ftheta\_camera\_rear\_left\_70fov | .mp4 | Video captured by rear left camera |
| ftheta\_camera\_rear\_right\_70fov | .mp4 | Video captured by rear right camera |
| pose | .tar | Camera poses derived from vehicle pose (30 FPS) |
| ftheta\_intrinsic | .tar | Camera intrinsic parameters for each view |
| metadata | .npz | Timestamp and pose information of the processed lidar data |

Camera and LiDAR Synchronization <br>
Camera Frame Rate: 30 FPS <br>
LiDAR Frame Rate: 10 FPS <br>
Synchronization: Each LiDAR frame corresponds to 3 consecutive camera frames. <br>
Pose Interpolation: Camera poses are interpolated at the starting timestamp of each LiDAR frame. <br>

Please visit https://github.com/nv-tlabs/cosmos-av-sample-toolkits to learn about how to use this dataset example.

## Dataset Quantification
Record Count: 10 synchronized clips of LiDAR and multi-view video <br>
Feature Count: LiDAR data, multi-view videos <br>
Measurement of Total Data Storage: 1.91 GB <br>

## Reference(s):
https://research.nvidia.com/labs/toronto-ai/cosmos_drive_dreams/

## Ethical Considerations:
NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.
Please report model quality, risk, security vulnerabilities or NVIDIA AI Concerns here.

# 数据集说明
本数据集包含10组激光雷达（LiDAR）与多视角视频的同步片段，并附带对应的相机与激光雷达位姿数据。本数据集可用于商业与非商业用途。

## 数据集所有者
英伟达公司（NVIDIA Corporation）

## 数据集创建日期
2025年10月23日

## 许可/使用条款
本数据集受[英伟达自动驾驶数据集许可协议（NVIDIA Autonomous Vehicle Dataset License Agreement）](https://huggingface.co/datasets/nvidia/Cosmos-Transfer-LidarGen-Example/blob/main/LICENSE)约束。

## 预期用途
本数据集旨在演示并帮助用户理解、使用Cosmos-Transfer-LidarGEN模型。

## 数据集特征
### 数据采集方式
* 自动采集/传感器采集
### 标注方式
* 自动采集/传感器采集

## 数据集格式
模态：激光雷达（LiDAR）、相机内外参、多视角视频
格式分类：激光雷达点云、多视角视频

您将看到以下文件夹结构：
传感器数据
| 文件夹名称 | 文件格式 | 说明 |
| :--- | :--- | :--- |
| lidar | .tar | 无运动补偿的激光雷达点云（帧率10 FPS） |
| ftheta_camera_front_wide_120fov | .mp4 | 前广角相机采集的视频 |
| ftheta_camera_rear_left_70fov | .mp4 | 后左相机采集的视频 |
| ftheta_camera_rear_right_70fov | .mp4 | 后右相机采集的视频 |
| pose | .tar | 基于车辆位姿推导的相机位姿（帧率30 FPS） |
| ftheta_intrinsic | .tar | 各视角的相机内参 |
| metadata | .npz | 处理后的激光雷达数据的时间戳与位姿信息 |

### 相机与激光雷达同步说明
相机帧率：30 FPS
激光雷达帧率：10 FPS
同步规则：每帧激光雷达数据对应连续3帧相机数据
位姿插值：相机位姿在每帧激光雷达数据的起始时间戳处进行插值。

请访问 https://github.com/nv-tlabs/cosmos-av-sample-toolkits 了解本数据集示例的使用方法。

## 数据集量化
样本数量：10组激光雷达与多视角视频的同步片段
特征类型：激光雷达数据、多视角视频
总数据存储量：1.91 GB

## 参考资料
https://research.nvidia.com/labs/toronto-ai/cosmos_drive_dreams/

## 伦理考量
英伟达（NVIDIA）认为，可信人工智能（Trustworthy AI）是一项共同责任，我们已建立相关政策与实践规范，以支持各类人工智能应用的开发。开发者在按照服务条款下载或使用本数据集时，应与内部模型团队协作，确保本模型符合相关行业与应用场景的要求，并应对可能出现的产品误用问题。
请在此处报告模型质量、风险、安全漏洞或英伟达人工智能相关问题。