dgjohnson/cvtest

Prototype version of Automap for Pixelblaze

CVTest - Experimental Version of Automap for Pixelblaze

Copyright 2024 ZRanger1.

NOTE: This repository contains the last experimental OpenCV/Python version of Automap before I decided to go all in on machine learning and the phone app thing. It will not be regularly maintained and is provided as-is.

Automap uses a webcam to map LED positions in 2D. It works with the Pixelblaze to step sequentially through each LED, and generates a map file which can be imported into the Pixelblaze mapping tab.

This code implements some of the core Automap ideas:

• it is fairly robust with regard to lighting conditions, even if they change during a run
• it implements adaptive thresholding and contour detection for improved detection reliability
• it works reasonably well with a variety of diffusers and LED types (I've tested LEDs diffused with paper, cloth, glass beads, fake rabbit fur, etc.)
• if, for some reason, it can't find a pixel, it will mark it with coordinates [-1,-1] in the map file so you can find it easily.
• you can rotate the generated map file to match the physical orientation of your LEDs. (No quantization or snap-to-grid though.)

The Downside:

• You'll want to be comfortable with Python to get the most out of this. Some editing of parameters may improve your results in certain situations.
• It's slower than I'd like. This is due to very slow frame capture in OpenCV's VideoCapture implementation.
• UI. There's kind of a UI, but it's just enough to support my experiments. Nothing you'd consider polished.
• The camera must remain still during the mapping process. (The phone version won't require this.)
• if your computer has more than one camera, you'll need to edit the code to select the correct one.

Dependencies:

• Python 3.10 or later
• opencv-python
• pixelblaze-client

You'll also want to install the Automap.epe pattern (included in the Pixelblaze folder) on your Pixelblaze.

To Run:

• be sure your Pixelblaze is on, with the Automap pattern running
• run the script: python CVTest.py <Pixelblaze IP Address>
• follow the instructions on screen.
• in CVTest.py, set self.debug = True (around line 20 ) to see additional diagnostic windows and debug output.