Skip to content

Iris segmentation algorithm to compare CPU and GPU (CUDA) performance

Notifications You must be signed in to change notification settings

jmiseikis/IrisSegmentation-CUDA

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

10 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

GPU (CUDA) vs CPU Iris Segmentation Performance Comparison

Coursework for a class during my master studies. The goal was to compare the performance of a Hough transform based iris segmentation algorithm implemented on CPU and GPU using nVidia CUDA framework.

Dependencies:

  • CUDA (and CUDA-enabled nVidia graphics card)
  • OpenCV

Example input images are provided in the images folder.

The algorithm uses Hough transform to find adjust an image and find a circle of eye's pupil (defined maximum and minimum diameter, relative to the size of an input image) in the image and search for a larger one with approximately the same center to get the edge of eye's iris. Then iris could be segmented and used for biometric or any other purposes.

Results:

Results

Runtime comparison CPU vs GPU:

Runtime

Very low-end GPU was used for testing (nVidia GeForce G210 with 512 MB), the acceleration would be significantly higher with medium-high end GPU. Both CPU and GPU implementations were not highly optimised, rather used as a proof of concept, and optimisation would improve the runtime in both cases.

Slideshow of the presentation: http://www.slideshare.net/jmiseikis/cuda-based-iris-detection-based-on-hough-transform

About

Iris segmentation algorithm to compare CPU and GPU (CUDA) performance

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published