LeszekSwirski / pupiltracker

Pupil Tracker
http://www.cl.cam.ac.uk/research/rainbow/projects/pupiltracking/
MIT License
49 stars 31 forks source link

Fail to compile in windows 7 #2

Closed willard-yuan closed 8 years ago

willard-yuan commented 8 years ago

Hi Leszek,

I try to build the pupiltracker, and it throw the following error:

1>  utils.cpp
1>c:\program files (x86)\microsoft visual studio 10.0\vc\include\random(1604): error C2064: Entries are not calculated as a function of the parameter accepts 0

The environment is Windows 7 and VS 2010. I think the main reason lead to this is the int pupiltracker::random(int min, int max) and int pupiltracker::random(int min, int max, unsigned int seed). Would you tell me how to fix it?

LeszekSwirski commented 8 years ago

VS2010 might be a bit old, can you try with 2015? You can download the community edition for free.

willard-yuan commented 8 years ago

@LeszekSwirski I change it to OS X and it runs successfully. I have read your paper Robust real-time pupil tracking in highly off-axis images. I can fully understand the paper, but when I test it on a video or a picture, the result seems to be not good. The followings is a result I get:

screen shot 2015-11-15 at 16 13 48

then I change to a small eye image( I used a HOG with linear SVM to get the eye area), the result is as followings:

screen shot 2015-11-15 at 16 37 36

The fitted ellipse by RANSAC draw on the result image is smaller than the size of the gaze. I'm reading the code fully now. How do I tune the parameters to get a good result?

LeszekSwirski commented 8 years ago

Have you adjusted the minimum and maximum pupil radius?

willard-yuan commented 8 years ago

@LeszekSwirski I have adjusted the params.Radius_Min to 0 and params.Radius_Max to 50. Now it works great. Using the same image above, I get the result as follows:

screen shot 2015-11-15 at 17 09 01

I'm really interested in your research and found you have do a lot of work in gaze tracking and eye model fitting. I'm really appreciate for your help and your great works. Thanks!