Scribble-Datasets-for-Segmentation

Below you can find the links to download the scribble version of segmentation datasets.

Features of the Scribble datasets:

• The scribble annotations generated by the help of DAVIS-Interactive.
• Following the COCO video format.
• The datasets are generated for the Cluster2Former, but those can be used for other projects as well.
• Due to the operation of the Cluster2Former, there are background scribble annotations as well, with "category_id": None
• The area of the scribble annotations of the objects are ~0.53% of the original annotation area.

The scribble datasets:

• YouTubeVIS 2019 Scribble / (original YouTubeVIS 2019)
• YouTubeVIS 2021 Scribble / (original YouTubeVIS 2021)
• OVIS Scribble / (original OVIS)

Installation and usage

See installation and usage instructions.

Examples

YouTubeVIS

Original annotations:

Scribble annotations:

Original annotations:

Scribble annotations:

Original annotations:

Scribble annotations:

OVIS

Original annotations:

Scribble annotations:

Original annotations:

Scribble annotations:

Original annotations:

Scribble annotations:

Acknowledgement

The Converter use codes from the DAVIS-Interactive.