There are 3 types of annotations in MultiScan: semantic label annotation, semantic OBB annotation, articulation annotation.
Script annotation_visualize.py generates turntable videos for the 3 annotations motioned above, with an additional video for the textured mesh of the scan.
python annotation_visualize.py input_dir=<path to downloaded dataset folder contains the scans> output_dir=<path to output folder will contain turntable videos>
Add additional argument scenes='["scene_00021","scene_00022"]'
to generate visualizations for only these 2 unique scenes (by default, the script generates visualizations for all the downloaded scans). Or change parameters in the config file.
outputs
├── scene_xxxxx_xx
│ ├── scene_xxxxx_xx.articulation.mp4
│ ├── scene_xxxxx_xx.instance.mp4
│ ├── scene_xxxxx_xx.obb.mp4
│ └── scene_xxxxx_xx.textured.mp4
...