« Updated Photosynth of Jen's Rooms | Main | Creating spherical panoramas »
August 24, 2008
Photosynth or PhotoCobble?
I spent this weekend playing with Microsoft's much ballyhooed Photosynth program, the first prototype of an application that uses Microsoft's Seadragon rendering engine to display images on the web. Unfortunately, Photosynth is a very rough Beta at this point and seems to be heading off on the wrong path.
There are a lot of comments that the product needs the following:
* the ability to specify a starting point image
* the ability to give app hints on how the images connect
* the ability to link one synth to another synth
* a preview before upload
* an option to abort the upload if synth% is too low
* more reliable upload/publishing
* better error handling
* an option to view the log file
These are all much needed, no doubt. But the greatest let-down in my opinion is that the app doesn't make any attempt to construct a virtual 3d world from the images. All it really does is show a progressive series of flat images that have been photo-cobbled together.
What I wanted, (since you asked), was to have the software assemble the images into a 3D virtual world that the viewer could spin around in 360 degrees endlessly, and then zoom to any point for a closeup. (Like this, for instance.)
What we get is something much different than this.
I made a lot of different photosynths this weekend. I followed the directions in the video tutorial and in the Photosynth Photography Guide (the directions are markedly different). I tried standing in the center of the room and spinning 360 degrees, taking a frame every 15 degrees. I tried standing in the corners and blanketing the room from each corner, and also from the middle of the walls, as instructed.
My hope was that the software would then seamless stitch the images together and create a 3D virtual world where the viewer could fly though with a mouse like they were riding a flying carpet. I was, of course, in for a Microsoft scale of disappointment.
Instead, (if it works at all), then for every position where you stood and pivoted to shoot pictures, you now have inadvertently created a unique physical location that a Photosynth viewer must navigate to in order to see those images. Argh!
The PhotoCobble team has managed to drag all of the limitations of meatspace into virtual world.
There should not be a separate movement from one "shooting position" to the next. Zooming in on a photo and movement should be one-and-the-same. In the virtual world, it doesn't make sense to separate the two. When I zoom into a photo, it is (or should be) the exact same as moving in that direction. Then, when I turn to the right, I should see what is on my right. I should be able to spin 360 degrees, and not hit the end of the photostitched panorama.
I find walking from point to point awkward, counter-intuitive, and tedious.
The goal should be to stitch the images into a viewable model using something like QTVR (QuickTime Virtual Reality) . Until then, I'm going to look for something else. This Beta version of PhotoCobble is too aggravating to play with any further at this point.
Posted by Rob Kiser on August 24, 2008 at 10:42 AM
Comments