search

OpenDriveLab / Vista

A Generalizable World Model for Autonomous Driving
https://vista-demo.github.io
Apache License 2.0
360 stars 15 forks source link

Only test on nuscenes? #6

Open IranQin opened 2 weeks ago

IranQin commented 2 weeks ago

Can it be tested on any given AD video using text conditions?

Little-Podi commented 2 weeks ago

on any given AD video

Certainly. I am not sure what kind of video format you want to use, but the easiest way would be to extract the initial frame and set the --dataset argument of the sampling script as IMG.

using text conditions

Our model does not accept textual inputs, as text can be unintuitive and ambiguous when specifying actions in the task of autonomous driving.

IranQin commented 1 week ago

Could you provide a mapping between the command and category of [0,1,2,3] ?

Little-Podi commented 1 week ago

Hi, we translate trajectories into commands following ST-P3. Please refer to this issue https://github.com/OpenDriveLab/Vista/issues/3 for details.