The updated robotcar-dataset-sdk includes scripts and functions to generate the downloadable mp4 videos by working with the RGB imagery and label masks for visible/occluded road boundaries.

Outputs for the following code snippets are available as examples and are taken from one of the samples available for download.

Raw road boundaries

Raw road boundary masks only indicate that a road boundary is present in that part of the RGB image, whether or not it can be seen:

python play_road_boundary.py 2015-03-17-11-08-44_small --type uncurated --camera_id left --masks_id mask

Classified road boundaries

Classified road boundary masks are marked as visible or occluded:

python play_road_boundary.py 2015-03-17-11-08-44_small --type uncurated --camera_id left --masks_id mask_classified

Curation

These classifications are semi-automatically produced, with some human fine-tuning available in the curated set:

python play_road_boundary.py 2015-03-17-11-08-44_small --type curated --camera_id left --masks_id mask

python play_road_boundary.py 2015-03-17-11-08-44_small --type curated --camera_id left --masks_id mask_classified

Stereo

Each of these mask types are also available for the right sensor in our stereo camera setup:

python play_road_boundary.py 2015-03-17-11-08-44_small --type curated --camera_id right --masks_id mask

python play_road_boundary.py 2015-03-17-11-08-44_small --type curated --camera_id right --masks_id mask_classified

python play_road_boundary.py 2015-03-17-11-08-44_small --type uncurated --camera_id right --masks_id mask

python play_road_boundary.py 2015-03-17-11-08-44_small --type uncurated --camera_id right --masks_id mask_classified

Videos

You may also want to save the video as an .avi to an output folder by using the arguments:

 --save_video --save_dir /output/folder/for/avi