Ask Your Question

Are there any new ways to improve eye detection?

asked 2015-03-05 08:58:08 -0600

kim123 gravatar image

updated 2016-01-21 10:04:17 -0600

I have been working with the OpenCV 2.4.10 Java libraries and Raspbian on a Raspberry PI 2. I got the basic face detection and tracking working pretty well on the RPI 2 and decided to go for face recognition. I have been translating the code from Mastering OpenCV with Practical Computer Vision Projects into Java and trying to build an application first on my laptop before trying it on the RPI 2. The problem is that the face detection works well, but the eye detection that is key to preprocessing images isn't working. I have tried varying lots of different parameters and using different cascade files, but the best I have been able to get is one eye accurately detected out of 10 images. I do have to mention that I am using images of myself wearing glasses, because I always wear glasses. The eyeglasses file actually seems to perform worse that the left eye and right eye files.

Some of the supposedly detected eyes on my face do not even remotely resemble an eye. So, I am looking for suggestions on ways to improve the detection especially with glasses.

edit retag flag offensive close merge delete



Did you try the eyeglasses model for detecting your eyes? It is impressively more accurate when applying it on people wearing glasses!

StevenPuttemans gravatar imageStevenPuttemans ( 2015-03-06 02:11:53 -0600 )edit

1 answer

Sort by ยป oldest newest most voted

answered 2015-03-05 14:21:24 -0600

Here is a "recently" published article about "A literature survey on robust and efficient eye localization in real-life scenarios" (2013) by Tan et al. (yes, the same author of the famous article "Enhanced local texture feature sets for face recognition under difficult lighting conditions") - Xiaoyang Tan google schoolar

See here: - Download is slow

To complete this answer, I would like to add information from ''

  1. A novel approach for accurate localisation of the pupils in real time is proposed in this paper: Timm and Barth. Accurate eye centre localisation by means of gradients. In Proceedings of the Int. Conference on Computer Theory and Applications (VISAPP), volume 1, pages 125-130, Algarve, Portugal, 2011. INSTICC. (also see youtube video at Accurate eye center localisation for low-cost eye tracking) There is also a github repository that implements this paper: C++ implementation And a blog article

  2. You can use this paper also to locate the position of the eyes. Bolme, D.S. "Average of Synthetic Exact Filters", IEEE Conference on Computer Vision and Pattern Recognition, 2009. CVPR 2009. PDF, Source code

  3. Another approach is to use a facial landmark detector. It will return you the four corner points of the eyes, corner of mouth, center of nose, and center of face. It does however require you to give it a bounding box of the face. You can use the four corner points of the eyes to reduce the region of interesting in order to find de pupils. project page PDF

edit flag offensive delete link more



Nice paper citations!

StevenPuttemans gravatar imageStevenPuttemans ( 2015-03-06 02:11:21 -0600 )edit

Question Tools

1 follower


Asked: 2015-03-05 08:58:08 -0600

Seen: 2,789 times

Last updated: Mar 05 '15