eye detection using SVM and HOG
First, I used a support vector machine to do binary classification to divide the eye data and not the eye data.
And I created a detector by creating windows of various sizes.
In the given test picture, the eye was detected correctly, although it was a bit inaccurate.
But when I tried other pictures, I did not get good results.
It seems to be caused by a very small amount of data sets.(about 400 eye data and about 300 non-eye data)
It was difficult to get a dataset, so there was a limit when I made it myself.
I think support vectors machines will perform very well and will produce good results if the dataset is large enough.
Finally, I tried to make a simple neural net, but it seems that it does not work well because of lack of time. I'll update the results later.