Camille Utterback:
Danny Rozin wrote the 'TrackThemColors' video tracking xtra for Director (available from www.smoothware.com) -- which Mary's code uses. Danny deserves some kind of medal for making video tracking available to people who don't want to write their own C or Java code, and for encapsulating so many useful tracking functions into easy methods. I highly recommend it for anyone who can code in Lingo and is interested in playing around with video tracking. (Danny also currently has a show up at Bitforms gallery in New York (www.bitforms.com) where you can see his video tracking software put to some rather unusual results.)

Mary -- having taken the time to figure out your code, I want to go back and add simple comments for people, but short of that, here's my gloss on what's happening: Mary uses Danny's 'TrackBlobs' method, which returns a list of rectangles indicating areas in each incoming frame of video that are different from the previous frame. Mary figures out the area and center point of each rect, adds the area and center point to a list, and sorts the list based on the area of the rects. The areas of biggest change are now at the end of the list, and she chooses the center points of the three last (and biggest) areas as the locations of her three 'dots.'

Mary's dots are *not* based on a visual characteristic of the video imagery that is easy to see. (I don't think our eyes naturally track centers of areas of change? Edges certainly). Her dots are based on a definition, however, that is easy to comprehend -- 'the centers of the biggest three areas of change between one frame and the next.' This is interesting because it points to the power of any linguistic structure, including code, to simply describe what is difficult or impossible for us to see. Add to this what Golan points out in his comments to me about machines being quick and tireless -- the computer can search for points that fit the definition over and over again 30 frames per second in this case - and we can now show what was only possible to describe. In Mary's piece we *can* now see the 3 dots flitting around from frame to frame, even if we could never have visually discerned what the dots represent.

I would like to see Mary's dots linked over time as well - representing the trajectory of areas of difference between one frame and the next. I guess this would be a hybrid of Mary's and my code ala the Scott and Mark hybrid applet that Martin made.