With all my fancy Max MSP work my main stumbling block so far has been finding the on switch on the speaker system…. PATHETIC!

Here are two tester videos. I was unexpectedly joined by two people who heard it from in the building. They were intrigued and took a break from work to talk to me about it and walk around. Sorry for the video on the floor most of the time, I was enjoying the success too much!

(click to full view it)

Boy am I glad that this works!

Taking a live feed from Atrium A in Portland Square at 3 different resolutions, but the same motion settings, I found these results.

160X120

320X240

640X480

At 160X120 resolution there were no undesired blobs. It tracked most of the people well, but it did completely miss others out. Each person was tracked as a single blob, but some pairs were tracked as a single, rather than two separate blobs. This ran fastest on my machine with sound distribution.

At 320X240 each person was tracked individually, and everyone in the video was tracked. There were a few stray blobs, but they were consistent. I can simply select those blobs and not use them as controllers of the sound. My machine struggled slightly while other apps were running, but this shouldn’t be a major problem.

At 640X480 everyone was tracked individually, but there were often 2 blobs designated per person. The sizes of these blobs is unpredictable and hard to eliminate. There was also some slowdown on my machine when runnign this with the sound distribution patch.

I’ve tweaked my patch slightly and found some good settings for that particular room. I did this by recording footage of myself walking around, and then running motion tracking on the loop on repeat, incrementally changing the thresholds and constraints of the tracking system. This is my best result so far, and one I am very happy with. There is very little difference between running from a prerecorded video compared with a live feed.

Here is a sample video of me in front of my webcam moving around like an idiot. I reckon I can calibrate this better anyway, but being this close to the camera with quite varied lighting are poor conditions for motion tracking anyway, and the results are great! The circles with crosses are the blobs that my Max patch has assigned, and you’ll notice that they follow segments of my body very well, and more often than not they also survive clashing with others, so long as the clash is brief they overlap well. This is something I had just written off as a quirk that I would have to put down as a flaw in my project.

• Entirely single page. I want everything in my portfolio to be found easily from the index page. The only externals I will have are links to download my CV (which is recently updated by the way) and to this blog.
• Tagged Content. In my old portfolio you could only find out what type of work my content was when clicking on it. It still worked because when you clicked on it it did specify what my part in projects were and what skills were used, but I would much rather have a more obvious system. I’ll be tagging content with a set range of classes that I can then filter out using jQuery
• Full content on load. I want my index page to show all the content available on loading the page. This makes it easy for users to find my content quickly, and it means any person stumbling upon my portfolio will be able to instantly see a variety for the work I produce.
• Nav-filtered tags. I will use left-hand navigation to filter the content types. On loading the page you will see all my work, but if you would like to specifically see my web work for instance, a simple click of ‘web’ will hide all the content that doesn’t involve a web presence. The same with video, branding e.t.c. This allows for anyone to get to a specific aspect of my work with a single click, without only exclusively showing that type. It says ‘Hey I can do a crap load of stuff, but if you just want to see this thing, then here you go!’
• Fluidity. I want everything on my portfolio to flow intuitively, both in looks and usability. I’ve built it to have my content as the central focus with simple navigation and no unnecessary clicks. Full viewing of my content isn’t built-in yet, but the thumbnails adhere to a 3 column layout, and the full content will take up all 3 columns. The styling is consistent and complimentary.

View the wip now

So whilst I have actually been doing Uni work I have also been preparing my new and much improved Portfolio. I’ve had my old Flash one for over 2 years now, and I haven’t changed it in the slightest, simply because I’ve been too busy with everything else. Stuff like a year of producing web and video content for Absolute Radio and doing some pretty cool stuff in my final year.

I’ve got all this work that I’ve done that my portfolio just doesn’t show, so I’ve been building a new look, showing off some of my newly acquired skills and stylistic taste and I’m liking it so far. It’s currently a WIP so I have a holder on my index, but you can have a sneak peek as well:

Past

Present

Future

Oh, and I jazzed up my branding somewhat. I’ve kept it as one fluid movement throughout the type but made it  bolder and better I feel.

It’s not something I’ve been struggling with technically – Converting my 1-sound patch to multi is easy enough to do, but I’ve been reworking the way each sound is calculated so that the subpatch is modular, and duplicatable with minimal editing of info.

Whilst playing around I also found the pictslider object, which is what I was after all along but couldn’t find! I used to use 1 horizontal slider and one vertical slider to control x/y but a pictslider does both at the same time in a more fluid way.

Old:

if listening to a Wacom this technique worked, but if I wanted to control from in the browser it wouldn’t, as one slider is directly on top of the other with a transparent background. I can only control horizontal movement with this…

New:

With this I have a background image of where the speakers are located, and I can move the dot around, with separate x and y outputs.

This simple change has allowed me to start experimenting with using multiple sounds as it is now very quick and easy for me to move sounds in the space. It was easy before with a WACOM but that limited me to just one sound. I can now move multiples at ease with my mouse.

What program did you use to recreate the effects water would have on the buildings?
Did you write that?

The program is a 3D modelling program called Cinema 4D. I didn’t write the program, just used it to model the buildings and have them move and change colour with contact.

I’m assuming the map of the shapes is a map of the buildings…

Yes it is. Well, buildings and individual features -There is a car park, and some of the boxes represent enclosed spaces at the backs of buildings, and there is a garden as well.

And how much water power are you applying to each person?

The way it works is based on a system of rules I have set up (find them in full on my blog http://orrinward.wordpress.com), but basically I treat a single person as a single unit, and a single car as a single unit (I don’t take size as a consideration). There are “hit zones” for each of the objects that are touching a road, the other boxes only move when touched by other boxes. The hit zones are effectively the path (not the road in the middle) and when a person comes into contact with a hit zone it pushes the relevant box 30 units (of whatever measurement it is in Cinema 4D) away from the person, fitting into space where possible.

When a box hasn’t been contacted for a while it returns slowly to it’s last stationary position – not the original position on the lane. If a person moves a box left twice, it can only naturally return to the first movement, not all the way.

Because there were very few people..and they had a huge effect…
Did you mean to do it like that?

Yes they had a huge effect. I did this show that the effect over a long time could be shown without craploads of footage. It takes ages to map the boxes out and the people, and also a while to render. I compressed it into a short space of time. That short animation has 6 pages filled with hand written co-ordinates.

Every time someone comes onto the screen I record the time they enter the lane, and the time they exit it. I then record the order of boxes they activate. I then have to convert the time that the video player gives me into a number of frames for the animation. It’s roughly equivalent to doing a 3600 frame animation completely manually, and having to synchronise it with a video. It’s not 3600 individual frames, but I have a keyframe (change event) roughly every 15 frames in the animation, so it’s at least 250 individual frames.

Issuu:

http://issuu.com/orrinward/docs/orrin_idat306

Rapidshare:

http://rapidshare.com/files/378301045/orrin_ward_IDAT306_10062098.zip

Final Video on Youtube: