I tried to take the lessons I learned from my Mark I trial last year to make the device elegant, easy, intuitive and totally automatic.
Construction of the Digital Picture Window
Here are some photos from the construction. It's all pretty trivial stuff. I bought a large, elegant frame from a craft store for $35. I had to use a router to greatly expand the indent in the frame to accommodate the bezel of the monitor (which was quite a bit wider than the area of a canvas you'd want to hide). I then had to miter the frame's sides so that it would fit the display, it was about 40% too wide. And a hot glue gun put the frame back together long enough for me to use some steel L brackets to guarantee longevity. I then placed the monitor inside, shimmed it so it looked right and then used the hot glue gun to simulate tack welding the screen to the frame. Once I was sure everything was even I used massive amounts of hot glue to secure the panel to the frame. I love hot glue, it peels off well when you need it to, but holds like a m**********r otherwise.
The panel is connected to a beautiful little computer housed inside a Mini ITX case. I flirted with using a Raspberry Pi as the brains, but sadly its performance is nowhere near good enough for the inefficiencies of ta prototype (video hardware and webcam performance are very poor). I also played around with some of the mini Android PCs in a stick and I'm convinced that is the way to go beyond the prototype stage, the performance seems quite good, the software support seems good, and the price and form factor are unbeatable. Once I get all my code working in Windows I'll begin porting to Android via Xamarin.
Last year for Christmas I deployed the first version of this project. Using an abandoned Dell XPS M2010 and marrying it with off the shelf software and a little AutoHotKey scripting glue I was able to create a device which automatically connected two remote spaces via high definition video and webcam whenever someone was in the remote space.
Its implementation was pretty simple, I used some freeware webcam monitoring software that would launch a user-defined EXE whenever motion was detected in front of the webcam (the software let you adjust sensitivity, exclude regions, define operating times, etc.). The EXE that was launched was an AutoHotKey script I wrote that would trigger a Skype call to a predefined user (the other endpoint). Each endpoint was configured in Skype to automatically accept the call and start video. Ending the session required manually hanging up in Skype. Some slight tweaking was done to make sure the BIOS would wake the machine every morning and task scheduler ensured it would shut down every night.
And that was that! It fulfilled all the basic requirements but its bulky and user-interaction requiring form ensured its ultimate demise. The device was quite large, requiring considerable desk space. The end user would frequently need to close the device to sort out papers, move things, etc. and once closed it was easy to forget to re-open it. Also, the fact that the device would always connect when there was motion was discouraging. People like the option of privacy and this afforded very little. There was no easy way to give it privacy features without making people interact with it like it was a computer, which is what I was specifically trying to get away from.
After a few months of use it was abandoned.
I was absolutely convinced of the idea's validity, it just needed a better implementation.