For this week in Computational Cameras, I wanted to create augmented reality camera icons that directly linked to IP camera feeds that can be found currently on the floor. These feeds include:
184.108.40.206 [[Currently set up in the main lounge area]]
220.127.116.11 [[PComp room attached to the Camera stand near the laptop]]
18.104.22.168 [[PComp room ceiling, the only locked camera with the same Ent users and the same friendly password]]
http://itp.nyu.edu/thesis/audience.html [[Room 50 / the room used when Speakers come]]
The idea was that I could find/create a 3d model of a camera to place on the floor within the AR platform Junaio, and then when users clicked on the camera it would show them the nearby camera feed (since they can all be loaded on a mobile browser).
First and foremost, I have no idea why Junaio elected to utilize an obscure 14 year old video game standard for 3d object modeling. Despite attempting to use the “officially endorsed” python script for Blender, trying to find other converters, and exploring the other 3d formatting option (.obj), I was unable to create my own 3d models to use.This ended up being one of the more serious problem, with the other being an inability to display loaded .obj files in AR space.
Through the interface on GLUE based channels, Junaio only allows small .md2 objects, complete with texture. After searching for ways to create my own, I decided my project would still function if I used a 3d model of a security camera (or any other type). However, again since the .md2 format is 14 years old, no one is developing models for it any longer. This resulted in a significant lack of free 3d camera based models to draw upon. When I was able to find suitable models of office furniture (yes, not a camera but something more interesting than basic shapes) the texture files ended up being too high definition (since the Junaio GLUE channel required .png < 500kb). I was able to find a model of a waving flag, which I implemented over the “Hiro” pattern formerly used in our AR work. I have hyperlinked this waving flag to the IP camera near the laptop in the PComp/Comp cameras room, and I will demonstrate this for class tomorrow.
With the location based channel I created, as previously stated I was unable to implement a 3d object in space. I altered the GPS coordinate within the file in order to better place the file, but was still unable to render a model within the space of either ITP or my apartment. Every time it would appear that I was on the mark, or at least close enough, the program would give me conflicting information. Once even Junaio believed I was both right on top of the object while simultaneously telling me that it was 47 yards away. While this is more the fault of the GPS / being inside…it was still frustrating to not even be able to clearly locate my markers in order to test in person my 3d models and their rendering (or lack thereof).
I would like to continue this project, mainly because I hope to figure out the AR in order to create location based 3d animated graffiti…but it seems that if I only want one or a few creations of my own I might be better programming it myself…