I’m now with Institute for Datability Science, Osaka University, Japan.
It’s a new institute and might not be very ready yet, but I’m very excited to set things up (kind of) from scratch.
Drop by Suita Campus!
I’m now with Institute for Datability Science, Osaka University, Japan.
It’s a new institute and might not be very ready yet, but I’m very excited to set things up (kind of) from scratch.
Drop by Suita Campus!
Our paper will be included in International Conference on Multimedia Modeling 2017, which will be held in Iceland in January (sounds cold!).
This work is a part of our MSCORE Project.
In this paper, we use two Kinect sensors facing to each other with an instructor between them to capture her/his motion, and reenacts the motion on a display with the mirror metaphor. We demonstrated its usefulness through user study.
Our paper has been accepted to IEEE Transactions on Visualization and Computer Graphics!
The online version is available here.
This work is to remove AR markers in a live video stream for visual aesthetics, which can handle even non planar surfaces. I joined this project just for technical help, mainly on GPU implementation of, e.g., Poisson blending.
Our paper has been accepted to Multimedia Tools and Applications!
The online version is available here.
This paper is on how to automatically generate a video summary for a blog post. We use the main text of that blog post to control the content of the video summary, so that it well suits to the blog post.
Our paper has been accepted to ACCV 2016.
Thank you very much for all authors!
This work proposes a new video summarization approach that uses deep features of a video to get its semantics. By training a deep neural network with sentences, our deep features encode “sentence-level” semantics of the video, which boosts the performance of a standard, clustering-based video summarization approach.
See you at ACCV 2016 at Taipei!
Our paper has been accepted to 4th Workshop on Web-scale Vision and Social Media (VSM) in conjunction with ECCV 2016!
This paper is on video/text retrieval by text/video queries. Our approach uses LSTM to encode text as many other existing approaches, but our observation is that LSTM tends to forget about the detail in the text (It mixes up “typing the keyboard” and “playing the keyboard”). The main contribution of this paper is to fuse to text representation web images retrieved using the text as query, which can disambiguates text.
Looking forward to see you at the venue!
Updated:
Now we have arXiv preprint. Please find it at: arXiv:1608.02367
Our project was recognized as excellent research work and collaboration in Microsoft CORE Porject 11! Thank you very much, the Team!
Our papers have accepted to EURASIP Journal on Image and Video Processing and ICME 2016.
The journal paper is on human action recognition using depth video sequence, and the conference paper is its application to sport video summarization.
See you if you attend ICME 2016!
To new webpage with a better URL!
Some old posts are removed, though…