JRDB-Traj only include stationary robot?
Opened this issue · 2 comments
Hi! @SaeedSaadatnejad @yanggao2000
Thank you for your excellent work!
I would like to know whether the JRDB-traj only includes the stationary robot. I found the scenes preprocessed in the train_traj_extractor.py do not include scenes where the robot is in motion (For example clark-center-2019-02-28_0 are excluded). On the other hand, the scenes preprocessed in the test_traj_extractor.py seem to include scenes where the robot is moving (For example, nvidia-aud-2019-01-25_0).
Thanks in advance.
JRDB-Traj/train_traj_extractor.py
Lines 93 to 108 in c6f92b5
JRDB-Traj/test_traj_extractor.py
Lines 27 to 56 in c6f92b5
Hi thank you.
In training, we used only scenes with the stationary robot, finding them to be more manageable for the model to train on. However, if you wish to train on other scenes, you can easily do so by adding their names to the list in train_traj_extractor.py
Thank you for your reply!
In training, we used only scenes with the stationary robot, finding them to be more manageable for the model to train on.
That makes sense because the model has to predict the future trajectories of other people and the robot's ego motion at the same time in the scenes where the robot is moving. I asked because I thought the trajectory was represented by the scene-centered coordinate and could not find any code for coordinate transformation from robot-centered coordinate to scene-centered coordinate. I understand the model is trained on the robot-centered coordinate.