|
Introduction
=========
Visual data, including internet images/videos, personal images/videos and etc., has been increasing in a rapid speed in the past decades. This brings researchers sufficient amount of data both for various kinds of deep learning model training and testing, in a theoretical way.
However, there are still a couple of issues remain unsolved, which greatly hinder the development of related algorithms in real applications. When data collection is concerned, most of the real-world visual data might have privacy issues, while computer generated data are still not real enough even with the state-of-the-art algorithms. Concerning data management issues, annotating large scale data is still a challenging topic, not to mention the complex annotation requirements for difference tasks.
This workshop will focus on solutions for these issues, aiming at providing constructive ways to build well-annotated, reliable, large scale visual dataset that can serve multiple purposes for academia and industries.
Research Topics
============
* Photo-realistic data generation/synthesis
* Large scale image/video data annotation
* Privacy protection for image/video dataset
* Video structure analysis
* Video summary and storyboard
* Cross camera video data analysis
* Multimedia retrieval
|
