I am using the Rotrics arm to make an image/drawing modifier. My prompt is the robot answering “this is what you should add.” Some of the responses will be humorous (doodling on faces, arrows to move things, etc.) and others will be more constructive (parallel offsetting existing lines, adding doodles inspired by parts of the drawing, etc.).
Inspiration:
- The Useless Box: https://www.youtube.com/watch?v=aqAUmgE3WyM
- Art Sqool: https://artsqool.cool/
- Google Quick Draw: https://quickdraw.withgoogle.com/
To start, I am working on a robotic L.H.O.O.Q. machine. Duchamp’s original piece is a mustache on a print of the Mona Lisa.
Image source: https://en.wikipedia.org/wiki/L.H.O.O.Q.#/media/File:Marcel_Duchamp,_1919,_L.H.O.O.Q.jpg
I have so far completed the facial recognition and am working on the pen and surface location mapping. Then I will draw a mustache on every face in the frame using the Rotrics robot arm. This is how I will produce the class postcards as well. I also want to try putting an actual head into the frame.
The location tags are implemented with April Tags and the face tracking uses Media Pipe Face Mesh. Face Mesh seems to require proportionally large faces within the frame (optimized for selfie cameras), so I have to limit the camera frame to 640 px by 480 px. The April tags seem to get more accurate with larger resolution, but not too much more accurate at this range.
My current goal is mustaches, but time permitting I want to also expand it to draw other features on not just faces. For example detecting contours and drawing offsets, filling in shapes, etc.