VLN-CE
VLN-CE copied to clipboard
jacobkrantz
[RxR-Habitat] How to export a robot navigation video based on a baseline
Could you provide me with the code in the image above "Generate a video of a robot navigating with natural language guidance based on baseline"?
You can use config options to generate a video during evaluation: https://github.com/jacobkrantz/VLN-CE/blob/2375f186cd8a1224a5ff15efa4c0ac05edfb2d42/vlnce_baselines/config/default.py#L23-L24
If you want to update the way in which frames are produced, you can modify/replace the functions in https://github.com/jacobkrantz/VLN-CE/blob/master/habitat_extensions/utils.py
Custom elements in the above videos (arrows, additional text) were added using opencv utilities.
Hi @jacobkrantz what can be reason for not getting a robot navigating video even after make changes to the [default.py](vlnce_baselines/config/default.py) as follows:
_C.VIDEO_OPTION = ["disk"] # options: "disk", "tensorboard"
_C.VIDEO_DIR = "data/videos/debug"
Do we need to make any others changes? One more question, what's the difference between using 'disk' and 'tensorboard' and how can they be accessed?
Hi @jacobkrantz what can be reason for not getting a robot navigating video even after make changes to the [default.py](vlnce_baselines/config/default.py) as follows:
_C.VIDEO_OPTION = ["disk"] # options: "disk", "tensorboard" _C.VIDEO_DIR = "data/videos/debug"Do we need to make any others changes? One more question, what's the difference between using 'disk' and 'tensorboard' and how can they be accessed?
Solved it. Thank you!