A Crowdsourcing Tool for Video Annotation

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering

Title: "A Crowdsourcing Tool for Video Annotation"

by

Mr. Leung Yu Hin

Abstract

Annotated video data are needed for many video processing research
problems such as object tracking and activity recognition.  However, video
annotation is considerably more complex than image annotation, making it a
very labor-intensive and costly process.  This research project seeks to
explore a novel crowdsourcing approach through a game with a purpose
(GWAP) to get the crowd involved in annotating video sequences.  A central
research problem is to effectively aggregate the data collected from
annotators with a wide range of expertise and quality levels.  We present
an automatic learning algorithm for detecting high-quality annotations and
skillful annotators using both key-frame annotations and continuous
annotations collected from the game players.  Experimental results show
that the proposed algorithm can successfully detect low-quality
annotations and spammers and it can improve the quality of the aggregated
annotations.

Date            :  30 July 2013 (Tue)

Time            :  3:00pm to 4:00pm

Venue           :  3501 (lift 25-26)

Advisor         :  Prof. D.Y. Yeung

2nd Reader      :  Dr. Helen Shen