About the train/test protocol

[asw91666]

Hello!

Thank you for sharing this excellent dataset.

I would like to know more about the training/evaluation protocol for the DD-Pose dataset.

1. The dataset includes two images of the frontal face (i.e., driver-left-img and driver-right-img) as well as images of the driver cabin. Is it correct that the driver-left-img was used for both training and evaluation?

2. Could you please clarify if all test image frames were used for evaluation, or if the test data was sampled (e.g., using every 5th frame for evaluation)?

Thank you very much.