The all-new HI-SPACE

I wrote a bit about the HI-SPACE table and my motivation for this project here.

There's also a fun little paper describing the original HI-SPACE (and a fun application-- virtual air hockey!) here

Overview

How hard is it to build a tabletop display that uses hands as input devices today?

Equipment/Tools

• Raspberry Pi 3, Model B
  • External display
  • Wired keyboard/mouse
  • Good power supply! (Get the right one-- you're gonna need it!)
• 8 GB SD card
• Raspberry Pi Camera
• OpenCV
• Video projector (...eventually)

Step 1: Get the Raspberry Pi up and running

• Image the SD card with the Raspberry Pi OS (I used the basic desktop version)
• Set up the dev environment on the Raspberry Pi:
  • Install the following:
    • list coming...
• Set up virtualenv and get the Python requirements installed.

I played around for a bit with the camera and OpenCV, but ended up just collecting and saving some images and video to use to develop the algorithms off-device.

I also housed the Raspberry Pi + Camera in a small box for some stability. Kinda rough, but gets the job done for the time being.

Step 2: Capture some starting data

Goal: Capture some still images of the background, and some video of the background + a hand reaching into the space. This will be used in the next step to develop the hand tracking algorithms.

• TODO: Get the code off the PI and put it in the repo!

Step 3: Open CV!

Goal: Identify different approaches to find and track a hand in the video.

A few different approaches to try:

• OpenCV provides a background subtractor-- wonder how far that gets us?
• Try a simple "subtract a photo of the background from the video frame"

Next Steps

Not there yet, but here's what's coming:

• Collect more images/video for testing the algorithms
• Profile the algorithms for time/space complexity
• Evaluate whether the algorithms are fast enough, and if not what to do different