Kia ora everyone!
So, it was pointed out by a few people that I haven’t talked about my Masters in a while. So, here I will give people the complete overview of what I’ve been doing and a misty view of what I will do. I have been working while the world cup was on! Just, only in the afternoons…
So, I’m still looking at a lil humanoid robot Wakamaru. We have some more information about it now, as we sent a postdoc guy over to meet the engineers and play with the robot for a couple of days.
So, now we know things like
* Its eyes don’t actually have cameras behind them – one of the forehead holes is a camera and the other hole is a microphone.
* It can say things typed into the code, but it currently only has Japanese phonetics built in.
* It’s voice recognition is ok, but it doesn’t actually understand much of what’s said. It aparently understands english, but I wouldn’t be surprised if it used Japanese phonetics again.
* Its face recognition can recognise up to 10 faces – not great, but interesting.
* The arms can’t lift anything heavier than about 500g (a block of butter). I suppose thats a functionality they’re working on.
* The hands can’t actually grab anything – they’re mainly there for gestures. Great.
* The comments (things which explain what the code is doing) is all in Japanese, which is bad…
* It has heaps of safety features built into it, like backing off if it bumps anything even a little bit, and lifting its arms in the air if it hits anything too hard etc. It’s been designed with the elderly and disabled in mind, so it wants to keep them safe.
* It needs to be told where its allowed to go (any flat area) and where its not allowed to go (stairs, wet areas, etc …). We’d rather it could work this stuff out itself…
And more and more and more….
So, this robot is meant to be coming over by the end of the month, which is kool. Will send you pictures and keep you posted.
So, if I was to sum up what I’ve been doing, stuffing about is a good description. I have been implementing different algorithms, reading the odd research paper and trying out demos this *whole* time. That’s it, seriously.
Mostly, I’ve been looking at object recognition – taking an image and being able to identify what the objects are in the image. I haven’t been overly successful, but have been trying out lots of different algorithms, like the one that reguired about 1000 samples of the object (like a mug) and then 10 000 samples of what was not the object (like cats, a street, clouds etc…) and then the algorithm (Haar Training) learns the difference based on that. Didn’t actually get that one working…
The one I’m working on now (Generalised Hough Transform) learns the shape of an object based on its edges, then tries to recognise that shape again by finding the same combination of edges. Its not working right now, but I hope to have it sorted by the end of the week. It looks like it’ll be too slow for real time application, but we’ll see.
I’m also going to play around with an algorithm called Distance Transforms, which also finds the shape of an object by finding its ‘skeleton’ then growing out to the shape. Something like that…
But this object recognition stuff seems less important if the robot can’t pick the objects up (non-grippy hands), so i think once I’ve finished the algorithm I’m on, I’ll brainstorm a new ‘focus’ as such, perhaps focusing the project on the human-robot interactions (manner, speech, functions) and just do the required computer vision to complete those essential functions. The overall idea is that Om (my research partner) and I can develop a sample system that would assist the elderly and disabled in their homes. It won’t be complete, and it won’t so ‘everything’, but it’ll do enough that we can show the potential.
So, hopefully that wasn’t too boring, but that’s where I’m at. I’ve had a couple of full nights sleep, but tomorrow morning is the first semi-final (Germany vs Italy) at 5am, and the game could go either way. if i get to bed early enough, maybe I’ll just wake up at 5am, watch the game to 7.45am (or 8.30am if extra time0 then go to uni. I might have to be pretty keen to do that.
And Rob’s coming again this weekend. There’s the Hockey Ball on this Saturday night, and Rob’s coming over to accompany me. Should be a great night. He arrives on Friday and goes back on Tuesday. With at least one job interview while he’s over here. *fingers crossed*
Hope that’s put some people’s minds at ease about my work. I have been working!