Video scene detection is the task of dividing a video into scenes. A scene is defined as a sequence of semantically related and temporally adjacent shots depicting a high-level concept or story1. Video scene detection is an essential pre-processing stage for many video analysis tasks.

The Open Video Scene Detection (OVSD) dataset is an open dataset for the evaluation of video scene detection algorithms. The open-source nature of the videos in the dataset makes them ideal to be used by researchers both in academia and industry. Additionally, the ground truth scene annotations were created using the movie script of these videos, thus making the ground truth annotations objective and unbiased.

1. Rui, Yong, Thomas S. Huang, and Sharad Mehrotra. "Constructing table-of-content for videos." Multimedia systems 7.5 (1999): 359-368.



Currently the dataset consists of 6 open-source videos, and continued work is being put into enlarging the repository.
The videos in the repository are:

Elephants Dream

Elephants Dream

© Blender Foundation | orange.blender.org

Big Buck Bunny

Big Buck Bunny

© Blender Foundation | peach.blender.org

Sintel

Sintel

© Blender Foundation | durian.blender.org

Tears of Steel

Tears of Steel

© Blender Foundation | mango.blender.org

Cosmos Laundromat - First Cycle

Cosmos Laundromat - First Cycle

© Blender Foundation | gooseberry.blender.org

Valkaama

Valkaama

© Tim Baumann | valkaama.com