LED Cartography

Experimental LED mapping tools to use alongside Fadecandy.

Infer layout from physical structure using a camera
Create linear 16-bit PNG images representing the light contribution from each LED
Generates Fadecandy config files and Open Pixel Control layouts
Converts layout to an editable SVG
Merge edited SVG back into layout

Dependencies

Fadecandy fcserver running
A DSLR camera that's compatible with gphoto2. I'm using a Canon EOS Rebel T5i
On Mac OS, you'll want this mac-gphoto-enabler tool to reversibly disable Mac OS's built-in PTP driver.
gphoto2, for talking to your camera
dcraw and netpbm, for decoding RAW images and extracting thumbnails
OpenCV, for analyzing images
Node.js and some NPM modules listed in package.json

On Mac OS, you can install libgphoto2 and dcraw with Homebrew:

brew install gphoto2 dcraw opencv netpbm
npm install

Make the environment as dark as possible aside from the LEDs
Set up the camera
- Manual white balance, daylight
- Manual focus
- Manual exposure, 1/20 or slower
- Manual ISO 100
- RAW image format
- Images should be fairly underexposed, to show as much detail in the highlights as we can get
- fadecandy/examples/python/chase.py is useful for testing the exposure
Run photographer.js. It detects all attached Fadecandy boards, and takes a photo for each LED
- It uses the camera's thumbnail to detect when the image stops changing, i.e. a particular LED string is done
- Each LED is photographed in pseudorandom order, to decorrelate any environmental noise from LED position
- It's always safe to kill and restart this script, it picks up where it left off
Output files in various formats are now in the data directory!
To generate a simple 2D layout and fcserver config, run mapper-2d.js

Automatically find a good exposure / LED brightness
Exposure bracketing / HDR
Support for multiple cameras and 3D reconstruction
Use detailed per-LED photos for interesting image-based rendering
Ways to capture the same data using fewer photos.
- Proof of concept binary mapper in the FC examples
- Use liveview mode to quickly segment image into nonoverlapping groups of LEDs