an investigation in movement through image database, zach lieberman, 2004
Picturing family is an installation that was installed inside the elevator at Ars Electonica's Museum of the Future during the Ars 2004 festival. It presented an ever-changing and non-repeating moving image sequence where the elevator's height directly controlled a virtual camera's movement, zooming in and out of family images from 1900 to 2004. The theme of that year's festival was "timeshift," and the project presented a simple cinematic response, where the users of the elevator could dwell upon the relationships between images of people from different decades.
The elevator at the Ars Electronica Museum of the Future is quite unique. It features a rear projection floor surface the users stand on top of. For this project, the elevator controlled live software that generated animation on the fly from an image database. As the elevator rose, the images moved forward in time from 1900 to 2004, and as the elevator descended, the images moved backwards. The motivation was simple: since the software was running in real-time from a database, visitors would rarely, if ever, see the same image repeated, and a simple elevator ride would have the potential to be as enigmatic and voyeuristic as peering into the photo albums of strangers.
The elevator at the Ars Electronica Museum of the Future is quite unique. It features a rear projection floor surface the users stand on top of. For this project, the elevator controlled live software that generated animation on the fly from an image database. As the elevator rose, the images moved forward in time from 1900 to 2004, and as the elevator descended, the images moved backwards. The motivation was simple: since the software was running in real-time from a database, visitors would rarely, if ever, see the same image repeated, and a simple elevator ride would have the potential to be as enigmatic and voyeuristic as peering into the photo albums of strangers.
(please upgrade your flash player
or enable javascript to see flash video)
--- bypass the detection! if you wish.
or enable javascript to see flash video)
--- bypass the detection! if you wish.
installation overview:
veoh (flash): external
mov (5mb): popup | download (save target as)
mp4 (3mb): popup | download (save target as)
veoh (flash): external
mov (5mb): popup | download (save target as)
mp4 (3mb): popup | download (save target as)
The genesis of this project, picturing family, was an interesting discovery. If you search google images with the term "family" and a given year, say "1955," you wind up with a hundreds of images which appear to be family photographs that were taken during 1955.
For example here is "family 1910"
and "family 1970":
I was amazed at all of the details these images had, and how from decade to decade they tell similar stories about why we photograph ourselves. While the technology changes around us, human nature never changes.
I immediately wrote some perl scripts and downloaded as many images for each year (1900-2004) as I could. The script ran for several days and collected thousands of images for each year.some examples of the images collected:
1909, 1917, 1926, 1937, 1944, 1955, 1967, 1974
The theme of ars electronica in 2004 was "timeshift" and it seemed fitting to create a small work for the festival using these images to explore the idea of the passage of time and the essential truths of images.
My intent in this project was to find a way to move through these images cinematically. This focus was based on the fact that I had seen many projects that coupled image databases, often quite successfully, with the metaphor of collage. I was interested in an alternative, that would allow a viewer to not only to develop a more personal relationship with an image, but as well, with the people that inhabit the image.
In order to create a cinematic experience, I decided to uses faces as key points for camera motion. I ran the image database through a face detection algorithm found in intel's open-source computer vision library opencv (the algorithm was Haar face features detection). I then outputted, for every image, all of the faces (including plenty of false of positives) as an external xml file.
Every camera motion, through every image, was to or from a random face, so that even if images were repeating, the potential to see the same motions were extremely limited as the highlighted faces would constantly change.
I am including here the source code for the face detector, which is based on sample code from the opencv library. It uses the following externals libraries: paramio (for XML ouput), devil (for jpeg image input/output). This code is written for windows (but easily portable) and lets you choose an image as input and after detecting faces in the image, ouptuts an xml file and a duplicate image with detected faces marked. enjoy.
By way of concluding, I have to share that I have never really felt too strongly about the field of information visualization. While I had seen great projects done in this field, it never really felt like something I would be interested in exploring aesthetically. It was only when I read this essay the anti-sublime ideal in data art by Lev Manovich, that I understood why. For me, this project was an experimentation with a very different approach to information visualization, one which I could completely relate to, which essentially chooses to live in the data, as opposed to extracting patterns and information from it.
thanks to the always super ars electronica futurelab, especially christopher lindinger and peter freudling, as well as the AEC museum staff. Also to golan levin for valuable software discussion. And many thanks to theo watson for the long days of help and the creative rivalry.