Ego motion of the dataset

[os1a]

Hi,

Do you provide the ego-motion of the driving car with respect to some reference point in the world coordinate system?

This can be similar to other datasets like nuScenes and Argoverse.

Thanks in advance

[peisun1115]

Yes, each frame comes with a pose. https://github.com/waymo-research/waymo-open-dataset/blob/master/waymo_open_dataset/dataset.proto#L281

That pose defines the transform from vehicle frame to global frame (see our coordinate system definition on the website (https://waymo.com/open/data/).

[jiaxinxie97]

Hi,
frame.pose is a Transform object, how can I convert it to an array?

[jiaxinxie97]

waymo-open-dataset/waymo_open_dataset/utils/frame_utils.py

Line 108 in 60df55a

np.reshape(np.array(frame.pose.transform), [4, 4]))

Thanks, I find the solution.

[nnsriram97]

Is it possible to get ego motions in terms of GPS coordinates? is lat,long,h for ENU coordinate frame of reference available?

[peisun1115]

Why do you need that?

[nnsriram97]

I would like to use map information from OpenStreetMaps. Hence if I can convert these to GPS coordinates it'll be helpful

[peisun1115]

Ah, we do not provide map information yet. We have plans for map but no ETA for now. Stay tuned.

[KushalBKusram]

Ah, we do not provide map information yet. We have plans for map but no ETA for now. Stay tuned.

As of today, have you added map information to the dataset?

[peisun1115]

Not yet. But we do have some work in progress. Stay tuned.

[KushalBKusram]

Not yet. But we do have some work in progress. Stay tuned.

Thank you! I am working on a research project and having GPS coordinates of the frame is quite essential. Could you share an expected timeline? It shall help determine if this information will be available anytime before my deadline. Thanks!

[peisun1115]

we do not plan to provide GPS coordinates.

[KushalBKusram]

we do not plan to provide GPS coordinates.

I am sorry but I am not familiar with coordinate systems. But is there any way that I could infer the location of the frame using the pose or information that you plan on adding to the dataset soon? If so, an approximate timeline? Thanks!