Tracking silouhettes

this is my first vvvv project. i am working on an installation which projects silouhettes of people on a screen. the silouhettes are tiled with symbolic images that represent them… as the people interact their silouhettes exchange images… so basically i need some method to track and distinguish each persons silouhette and draw the correct images that correspond to that persons personality on the silouhettes. i read some posts related to tracking and it seems ill need to use freeframe? thanks.

right. have a look at the Contour (FreeFrame DShow9) node and its helppatch. that should get you startet.

Note that you won’t have an easy time finding out which silhouette belongs to whom once the people pass in front of each other. You may have to evaluate the colors of their clothes, too. Unless you want the images to exchange any direction whatsoever.

hi. atm ive been unsuccessful at drawing an image of the silouhettes from the contour module. im using help patch as a reference… i connected the x and y outputs to line. what would be the best way to draw the contour if i wnated to add a texture to the sillouhette?

ps im tried the vvvv irc channel and no users were around except @newk. im not familiar w/ irc so i mayve done something wrong and the one inside vvvv doesnt work. how else can i chat w/ other vvvv users? im coming strictly from a textual programming backround and am determined to learn and understand this to the best of my knowledge so any help is appreciated.

i recently found the project jannis did. and its an good example for what i want the output of my program to be.

http://weblogs.digital.udk-berlin.de/jannis/archives/002446.html

except i dont need to project images on peoples bodys. i just want to take their body sillouhette and apply a simple static texture to it. it doesnt seem like he used contour or camshifttracker.

hey my friend,
i’m currently on a little jam in paris but i will be back home in berlin
tomorrow night.
sure thing, i’ll do my very best to help you with your problem. indeed i
didn’t use the contour node, it wasn’t available when i did the body projections for “Generative Audiovisual Systems”.
I wanted to put all my patches online a while ago but things keep my busy…
I will check out my old patches and make something for you…
will put it out for everyone as well.
It’s a little tricky the whole thing, you need an infrared suitable camera.
A standart handycam with nightshot works fine. you might need an infrared
filter as well. some simple pixelshader etc…
more tomorrow when i’m back home…

I FINALLY UPLOADED A LITTLE ONE MINUTE TRAILER OF THE SHOW showing body projections and computational beauty in action!!!

http://www.DeinLieblingsgestalter.de

cheers jannis

thanks jannis. im really excited!

hype:
ps im tried the vvvv irc channel and no users were around except @newk. im not familiar w/ irc so i mayve done something wrong and the one inside vvvv doesnt work. how else can i chat w/ other vvvv users? im coming strictly from a textual programming backround and am determined to learn and understand this to the best of my knowledge so any help is appreciated.

i’m not really a helpdesk for vvvv… i’m actually a total vvvv beginner myself. But i hang out allot in vvvv just to share some info and ideas, if you are connected permanent like me you can stick around in the irc channel if you want.
the more people hang out there the more direct response could be given.
cheers

from our mail correspondence:

to answer your question:
yes i did use tracking ! the content in the masks follows the moving artists at all times.

i customized the visuals for each song that was performed. for the song shown on the images i projected the lyrics on the rappers.

http://weblogs.digital.udk-berlin.de/jannis/archives/002885.html

the body is literally a container, carrying their thoughts.
it empties in realtime while the artist is rapping. the lyrics stream through the body to the mouth where they leave the body. at the end of the song the body is empty. you can see a very short sample of this in my trailer.

http://www.DeinLieblingsgestalter.de

if you watch closely you can see how the bodytexture travels with the moving artist. to patch this in vvvv is not too hard…what you will experience in your design process is how difficult it will be to illuminate the stage evenly with infrared light until the tracking works properly.
the identification of the rappers is needed only for some songs.
a fast solution to solve this is to simply split the stage in 2 parts. until each one stays on one side it’s not a problem to project different content. it is possible to identify persons even if they move free across the stage and stand behind each other.

this is way more sophisticated, i do have the knowledge how to do that. it was neither in my time nor in my money budget.
more tomorrow…

cheers from berlin, jannis

The saga begins: Workshop Day 1
Here is the first part of the “How to: Body Projections” workshop by Jannis Kreft.

Today we start it simple, introducing principles of blend modes and layering to do some masking and we use a pixelshader to do some vvvvodoo. The result of day 1 will be a texture projected only onto a body (not the background). The texture can be a video or an image. Later we will learn how to project computational realtime beauty instead of a video.
For more insightful info visit Jannis at http://www.DeinLieblingsgestalter.de or simply enter the dragon and find your way to Berlin(West), home of the free, home of the brave.
1000 kisses, peace, I’m out.

DLG_BodyProjections.rar (9.5 kB)

yay… no text …