ESP for computer vision tasks

[nebgnahz]

It's probably quite complicated to directly support OpenCV inside ESP (although ofxOpenCV exists). But since we have a richer set of input streams (like TCP #364, OSC #365), we can wrap some intermediate detected results from OpenCV and then use ESP to extract gestures or other inference.

One proof-of-concept is available at: https://github.com/nebgnahz/esp-vision.

Note: to run it, the application right now depends on a Rust binding for OpenCV (which is currently in a private repo but will eventually be open-sourced). Setup the ESP gesture recognizer using TCP input stream TcpInputStream stream(8001, 2), this needs to be started first. Then run the computer vision application esp-vision. Use mouse to select a region to track and the tracking centroid is sent to ESP (space separated x and y values).