Felipe Codevilla

Hey, this was hosted at the CARLA google drive. I am checking what happened there.

There are several reasons. Listing some: - You are using a very different CARLA version. CARLA changes considerably the pedestrians and car models and that can affect the results. The...

Yes, I found this recently. It was dividing twice. Thanks for pointing this out. I wouldnt' change that on the repository since this is necessary in order to reproduce the...

Absolutely. THe only problem would be the driving module from the system. But there will be no problem on training and validation !

I not sure I understand the issue. The pre-trained models are made using 0.8.4 data, so they probably wont handle scenarios from 0.9.X. The data collector is still not updated...

Basically you just need images and json files with steering/gas/brake/speed. Those can be easily obtained with the scenario runner.

That message is quite bad haha ! Try running train only to get the full message. add to the options --single-process train -e coil_icra

I have just checked it seem to be fine.

This should be fixed, I was using an older version of the scenario runner.

+1 for this