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
485 stars 67 forks source link

Oflfine tracking limited to 15-16FPS, any way to improve it? #33

Closed catubc closed 3 years ago

catubc commented 3 years ago

Hi We are starting to use your tracker, and are finding that 15-16FPS is the fastest that it processes our videos (acquired at 30-90Fps). Here's an image of running eyeloop for a previously recorded .avi.

https://drive.google.com/file/d/1UrOGWEx92dEnImPnksNqxVUA9gKDin0N/view?usp=sharing

We were thinking about digging into your algorithms to speed them up, but I assume you guys know better than us how to do this.

Any advice for this?

For closed loop (online) tracking this would generally not be enough.

Thanks so much, catubc

simonarvin commented 3 years ago

Hi Cat, perhaps you haven't initiated tracking (you're still left in a slowed-down stream for selecting the pupil and/or corneal reflections).

After selecting the pupil, (press z, then y). This initiates tracking at full speed.

catubc commented 3 years ago

Yes, you right, sorry didn't see the info in the docs. I see it now.