Whether the EEF action in the dataset has been normalized

[zhang2514yuchi]

I want to use the model trained by Droid dataset to do inference in the real environment. If there is a regularization operation, we may need to reverse the regularization.
I couldn't find any information about the droid datasets. I want to know if the raw data is normalized, and if so, how it is normalized.

[kpertsch]

Hi!
The data is not normalized -- if you fit a policy to the actions in the RLDS dataset or the raw h5 files, it should replay on the DROID hardware setup.
The being said, during our policy training experiments we did apply normalization (but it's not part of the data itself).

[zhang2514yuchi]

Thanks for the explanation, I have another question: Are the angles of the end effector recorded in Euler angles?

[AlexanderKhazatsky]

Yes they are recorded in Euler angles!