TubeFiler - an Automatic Web Video Catgorizer
In context of the
Multimedia Grand Challenge, a set of open problems specified
by industrial partners, we present the
TubeFiler framework, as our contribution to the
Google Challenge: Robust, As-Accurate-As-Human Genre Classification
for Video. The Google Challenge particularly
aims to encourage more work in the area of semantic understanding of a broad variety of videos.
The
TubeFiler framework provides two key features:
- an automatic multimodal categorization of videos into a pre-defined genre hierarchy
- the support of additional fine-grained hierarchy levels based on unsupervised learning
For each genre, videos are downloaded from
YouTube together with their tags and titles. Both visual features and metadata are used to train a
statistical model, which is then used to categorize new incoming videos.
Reference
TubeFiler - an Automatic Web Video Categorizer
Damian Borth, Joern Hees, Markus Koch, Adrian Ulges, Christian Schulze, Thomas Breuel, Roberto Paredes
ACM Multimedia Grand Challenge, 2009
