Scott

Yes, the motion dataset contains multiple examples (scenes) per record. We do not currently provide a way to determine the type of scene or location for the motion data. We...

Hi, The issue is that the labels are in the local vehicle frame coordinates but the map data is in global frame coordinates. For each frame you need to use...

The pose field as used in the code above is a standard 4x4 transformation matrix. You can use it to either transform a unit 3 dimensional vector representing the heading...

Something may be going wrong with your transformation of the trajectories. The example code [here](https://github.com/waymo-research/waymo-open-dataset/issues/736) shows how to do this transformation. In general, you should be able to align the...

You should submit with all of the scenarios in the dataset. The server has a very small allowance for missing scenarios. This will likely be removed in near term updates....

The map data should exist in the first Frame proto in each sequence. It only exists in the first frame as the map data is constant for all frames. The...

There is not a correspondence between the perception and motion datasets. They are generated independently and are distinct data. In addition, the motion dataset is over 100X larger in terms...

This was an issue on our server. Could you try it again? We believe it has been resolved.

If you are referring to the motion dataset, unfortunately we are not able to provide absolute world coordinates at this time.

Hi, The two versions of the datasets have exactly the same set of scenario IDs, however they are sharded differently, i.e. the file uncompressed_scenario_training_training.tfrecord-00000-of-01000 will not contain the same subset...