What am I doing in Japan, really?

By now, loads of people have asked me what the **** I'm actually doing in Japan. As if karaoke, parties and food poisoning is not enough. So, well, I guess I'd better tell y'all.

The short answer is I'm working on my Master's Thesis. Yes, I was supposed to do that years ago. For several reasons, that never happened. So there I was, last winter, after three years working, still one thesis short of a Master's degree. It was getting kind of silly.

Last January me, Richard and Stef went on vacation in Japan. While there (well, here), we met Mats, who was writing his thesis at KDDI. Matter of fact, if I understood things right, it was his second thesis... So I figured, heck, if he can write two, I'll be darned if I can't get to one. I applied through Uppsala University for the same position Mats had had, long story short, I got it.

What I'm doing here is mostly related to video processing. KDDI Research Labs (which is where I work at) already has, largely thanks to Mats, a system that can do some tracking of video news. I'm trying to improve that system by adding and exploring new capabilities.

At this point, some of you will be nodding in agreement, aha, OK, tracking video news, got it. But most will be wondering "OK, now, what the heck does that mean then!?".

Have no fear. I'll try to explain.

In the future, people will be getting video news feed or clips through internet, mobile-phone, TV and maybe other, unforeseen, channels. There will not be fewer things to look at than there are today, in fact, most likely there will be more. Also, more and more people acquire the ability to record these clips - through video recorders with hard drives, computers, what not.

But here's the problem - the more stuff you got stored, the harder it will be to find something. And as more news is thrown at you, it gets harder to pick out the pieces that matters to you.

This is something that a lot of researchers has realized. There are many teams working on solving different parts of this. The future common though is a system where a user can ask the computer for certain video clips, or images, without having to apply "tags" or anything else in advance. A system where the computer, in one way or another, "understands" what's in the video clips and images.

(I put understands in quotation marks because I'm using it in a very loose meaning ... I do not intend to get into a philosophical question on whether computers can indeed understand anything at all).

Our niche, so to speak, is to focus on video images and news only, push that as far as possible. Given one news story, tell me when more is shown - that is in some ways the most basic question that we're trying to answer. We don't worry about speech or on-screen text - many other people are doing that. We're merely trying to push image and video recognition as far as possible.

And as for me, personally, I'm just trying to add one or two more bricks to what might one day become a nice big house.


Something else: In a few days, I'll post a write-up about my recent trip to Kyoto, along with pictures of some absolutely stunning temples and sights... So - pics are on their way! :-)