search

trishume / eyeLike

A webcam based pupil tracking implementation.
MIT License
917 stars 335 forks source link

Is it possible to integrate your eye-center implementation to opencv-contrib? #7

Open comdiv opened 8 years ago

comdiv commented 8 years ago

Today our team works hard with opencv "face" module (hosted on https://github.com/Itseez/opencv_contrib). So when it's possible and usable we try to contribute to it while we think - it's better to keep core logic of CV in libs that close to OpenCV itself and managed with big comunity. For some reasones we try to use your eye-center implementation and it's looking as simple and effective one. If you are not disagree we think to copy and rewrie your code and then provide as pull request to https://github.com/Itseez/opencv_contrib as a part of "face" module. You havn't apply any license to your repository. So i think we should apply usual opencv license to it and then write after it "Initially based on https://github.com/trishume/eyeLike implementation"

trishume commented 8 years ago

Sounds good to me.

If you're interested there's also a not-that-difficult optimization that might lead to a 100x performance increase. Right now it is inefficiently O(n^2) in number of pixels and thus O(n^4) in width. It could be O(n) in pixels by using gradient ascent, as detailed in Dr. Timm's thesis. That's the main quality issue with this implementation. It's possible I'll end up implementing it this winter, but not totally likely.

This repo really should have a licence, I'll try and put it under MIT soon and since I wrote all the code I can relicense it under another license if necessary to put it in OpenCV contrib.

trishume commented 8 years ago

Derp apparently I put it under MIT years ago, which is good, I was worried that I had forgot to licence this project and then people actually used it.

comdiv commented 8 years ago

When we setup some sources we will primary show it to you to check license.

DWCTOD commented 3 years ago

Sounds good to me.

If you're interested there's also a not-that-difficult optimization that might lead to a 100x performance increase. Right now it is inefficiently O(n^2) in number of pixels and thus O(n^4) in width. It could be O(n) in pixels by using gradient ascent, as detailed in Dr. Timm's thesis. That's the main quality issue with this implementation. It's possible I'll end up implementing it this winter, but not totally likely.

This repo really should have a licence, I'll try and put it under MIT soon and since I wrote all the code I can relicense it under another license if necessary to put it in OpenCV contrib.

hi, sir, thanks for your nice work, could you tell me now eye-center implement is integrate to opencv-contrib? Looking forward to your replay , thanks you very much.