simonarvin / eyeloop

EyeLoop is a Python 3-based eye-tracker tailored specifically to dynamic, closed-loop experiments on consumer-grade hardware.
GNU General Public License v3.0
479 stars 68 forks source link

improve blink detection #12

Closed simonarvin closed 3 years ago

simonarvin commented 3 years ago

Background

EyeLoop detects blinking by monitoring changes in the monochromatic spectrum (= changes in the mean brightness of the frame).

It turns out that blinks produce very distinct patterns in the mean value when viewed over time, which can be used to detect blinking. How to threshold this pattern depends on the species monitored (eg human, mice) and the general properties of the image (eg contrasts). Thus, we have to normalise this function to improve detection across species and footage types.

Get started

check_blink() is found in /eyeloop/engine/engine.py. For human and mice test footage, click here.

Best, Simon

tosimplicity commented 3 years ago

It would be even amazing if we can detect intentional blinks (vs involuntary eye blinks).

simonarvin commented 3 years ago

see #24