metadriverse / aco Goto Github PK

Thanks for sharing your great work! Just wonder is it possible to share your zip file of those sampled frames from the youtube videos? It is a little bit time and space consuming to download all the videos.
Thank you in advance.

Hi ACO Team,

I wanted to ask that do you guys have the information regarding the camera intrinsic matrix for the videos present in the YouTube driving videos dataset?

Thank you.

HI, I have several questions about the model used for PPO training.

• Does the model takes navigation command and speed as input the same way as CILRS?
• Do the actor encoder and the critic encoder share the same model?

Hello, this is Yuan, who are interested in this great work! I wonder which dataset is the provided pre-trained weight trained on and what the corresponding sampling rate? Thanks!

Hi, how many training samples are used for the training of the Inverse dynamics model (only samples data or all the sweeps data), and do you scale and center crop the nuscenes image to 320*180 for training?
Also, when calculating the optical flow for training, is the time interval between current and next image 1s?
Thanks in advance.

Hello ACO team,

This is Yuan from UIUC. I wonder if it is possible to share your imitation learning model and pre-trained weights with me at [email protected]? We hope to use it in our ongoing project if possible. Thanks a lot for your help!

Yuan