Skip to content

about Coordinate System #18

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
Hanxuan411 opened this issue Apr 2, 2025 · 6 comments
Open

about Coordinate System #18

Hanxuan411 opened this issue Apr 2, 2025 · 6 comments

Comments

@Hanxuan411
Copy link

Hello, I would like to learn about the work related to the coordinate system of this project, such as the selection of the world coordinate system, the definition and conversion between coordinate systems, etc., to help me gain a deeper understanding of this project.
I look forward to your reply!

@wswdx
Copy link
Collaborator

wswdx commented Apr 2, 2025

Hello, I would like to learn about the work related to the coordinate system of this project, such as the selection of the world coordinate system, the definition and conversion between coordinate systems, etc., to help me gain a deeper understanding of this project. I look forward to your reply!

Hi! In our code, the world coordinate is set to be the same as LiDAR frame in nuScenes, which is in RFU (x-right, y-front, z-up) convention. The camera coordinate is set to be the same as camera frame in nuScenes, which is RDF (x-right, y-down, z-front). For each bin in our data, (0,0,0) of the world is set to be LiDAR position of the central frame.
You can refer to https://www.nuscenes.org/nuscenes#data-collection for more details about nuScenes coordinate system.

@Hanxuan411
Copy link
Author

I understand, thank you for your reply!

@Hanxuan411
Copy link
Author

您好,我想了解一下这个项目坐标系的相关工作,比如世界坐标系的选择、坐标系之间的定义和转换等,以帮助我对这个项目有更深入的了解。我期待您的回复!

你好!在我们的代码中,世界坐标设置为 nuScenes 中的 LiDAR 帧相同,这是 RFU(x-right、y-front、z-up)约定。摄像机坐标设置为与 nuScenes 中的摄像机帧相同,即 RDF(x-right, y-down, z-front)。对于我们数据中的每个 bin,世界的 (0,0,0) 被设置为中心帧的 LiDAR 位置。您可以参考 https://www.nuscenes.org/nuscenes#data-collection 了解有关 nuScenes 坐标系的更多详细信息。

I'm sorry to bother you again. Are you saying that this project did not use ego vehicle frames, and the data was processed to use lidar frames as global frames? May I ask if the lidar frame in your project can be consistent with the lidar frame in the nuScenes dataset, and if the points in both can be directly merged?
I look forward to your reply!

@wswdx
Copy link
Collaborator

wswdx commented Apr 3, 2025

您好,我想了解一下这个项目坐标系的相关工作,比如世界坐标系的选择、坐标系之间的定义和转换等,以帮助我对这个项目有更深入的了解。我期待您的回复!

你好!在我们的代码中,世界坐标设置为 nuScenes 中的 LiDAR 帧相同,这是 RFU(x-right、y-front、z-up)约定。摄像机坐标设置为与 nuScenes 中的摄像机帧相同,即 RDF(x-right, y-down, z-front)。对于我们数据中的每个 bin,世界的 (0,0,0) 被设置为中心帧的 LiDAR 位置。您可以参考 https://www.nuscenes.org/nuscenes#data-collection 了解有关 nuScenes 坐标系的更多详细信息。

I'm sorry to bother you again. Are you saying that this project did not use ego vehicle frames, and the data was processed to use lidar frames as global frames? May I ask if the lidar frame in your project can be consistent with the lidar frame in the nuScenes dataset, and if the points in both can be directly merged? I look forward to your reply!

Yes, the LiDAR frame used in this project is consistent with that of nuScenes. And yes, theoretically, the LiDAR points in nuScenes can merged into ours. We do this because using LiDAR frame as world frame is a common practice for most BEV perception methods in the field of autonomous driving. We just follow them and use the same convention.

@Hanxuan411
Copy link
Author

您好,我想了解一下这个项目坐标系的相关工作,比如世界坐标系的选择、坐标系之间的定义和转换等,以帮助我对这个项目有更深入的了解。我期待您的回复!

你好!在我们的代码中,世界坐标设置为 nuScenes 中的 LiDAR 帧相同,这是 RFU(x-right、y-front、z-up)约定。摄像机坐标设置为与 nuScenes 中的摄像机帧相同,即 RDF(x-right, y-down, z-front)。对于我们数据中的每个 bin,世界的 (0,0,0) 被设置为中心帧的 LiDAR 位置。您可以参考 https://www.nuscenes.org/nuscenes#data-collection 了解有关 nuScenes 坐标系的更多详细信息。

很抱歉再次打扰您。你是说这个项目没有使用自我车辆框架,数据处理后使用激光雷达框架作为全局框架?请问您项目中的激光雷达帧是否可以与 nuScenes 数据集中的激光雷达帧保持一致,以及两者中的点是否可以直接合并?我期待您的回复!

是的,本项目中使用的 LiDAR 帧与 nuScenes 的帧一致。是的,理论上,nuScenes 中的 LiDAR 点可以合并到我们的点中。我们这样做是因为使用 LiDAR 框架作为世界框架是自动驾驶领域大多数 BEV 感知方法的常见做法。我们只是遵循它们并使用相同的约定。

Thank you for your reply. At the same time, I found that the generation of volume_gs in your project can input pixel-gs and features within the voxel grid range. Is this necessary for better results? Thank you again for your reply!

@wswdx
Copy link
Collaborator

wswdx commented Apr 8, 2025

您好,我想了解一下这个项目坐标系的相关工作,比如世界坐标系的选择、坐标系之间的定义和转换等,以帮助我对这个项目有更深入的了解。我期待您的回复!

你好!在我们的代码中,世界坐标设置为 nuScenes 中的 LiDAR 帧相同,这是 RFU(x-right、y-front、z-up)约定。摄像机坐标设置为与 nuScenes 中的摄像机帧相同,即 RDF(x-right, y-down, z-front)。对于我们数据中的每个 bin,世界的 (0,0,0) 被设置为中心帧的 LiDAR 位置。您可以参考 https://www.nuscenes.org/nuscenes#data-collection 了解有关 nuScenes 坐标系的更多详细信息。

很抱歉再次打扰您。你是说这个项目没有使用自我车辆框架,数据处理后使用激光雷达框架作为全局框架?请问您项目中的激光雷达帧是否可以与 nuScenes 数据集中的激光雷达帧保持一致,以及两者中的点是否可以直接合并?我期待您的回复!

是的,本项目中使用的 LiDAR 帧与 nuScenes 的帧一致。是的,理论上,nuScenes 中的 LiDAR 点可以合并到我们的点中。我们这样做是因为使用 LiDAR 框架作为世界框架是自动驾驶领域大多数 BEV 感知方法的常见做法。我们只是遵循它们并使用相同的约定。

Thank you for your reply. At the same time, I found that the generation of volume_gs in your project can input pixel-gs and features within the voxel grid range. Is this necessary for better results? Thank you again for your reply!

Yes. It's suggest to include this projection step for reproducing our results.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants