Example

In the Camera2DVD project we combine state-of-the-art visual and audio analysis techniques to automatically edit raw home video material (e.g. from DVCams or mobile phones) to entertaining videos. The idea is similar to video summarisation or video skimming where hours of raw video data is "compressed" to a few minutes of more enjoyable but almost equally informative video.

The ultimate application of our research would be a (free) service, where users can upload hours of raw video material and after automatic processing they would receive edited home movies where raw material is skimmed to entertaining movie of requested length.

The project has been funded by the Town of Kouvola, Finland, and European Union Regional Development Fund. We collaborate with the Audio-Video Production team in Kymenlaakso University of Applied Sciences (KyAMK)

Town of Kouvola EU Regional Funding Kymenlaakso University of Applied Sciences

People

Antti Hietanen E-mail Research Student
Jukka Lankinen E-mail Post-doc researcher
Katariina Mahkonen E-mail Research Student
Jori Pölkki E-mail Senior Lecturer (KyAMK)
Tuomas Virtanen E-mail Research Fellow (Audio Group)
Joni Kamarainen E-mail Professor

Selected Publications

Video Shot Boundary Detection Using Visual Bag-of-Words

By Jukka Lankinen and Joni-Kristian Kamarainen In Int. Conf. on Computer Vision Theory and Applications (VISAPP) 2013.

Our method detects shot boundaries automatically using the popular approach in visual object class detection: visual Bag-of-Words. In particular, colour (RGB) and BoW histograms form a complementary set of features that provide excellent results via simple processing pipeline.

Download: PDF, BibTexSource Code (Bitbucket)

Links

Video editing tools

Kdenlive
OpenShot
Cinelerra
Pitivi
Linux Journal Editor's choice 2013

Data sets

TrecVid
The most popular evaluation dataset for video processing

Authors and teams

Yael Pritch
Video synopsis research
Ba Tu Truong
Video synopsis research
John R. Kender
Analysis of various video types