Dear All,
Good Day,
Im using facial features for an application. I am required to use a fast face detection methods/algorithms.
At present im doing my implementation on MATLAB and using viola jones face detector. Im getting a frame rate of 3 fps.
can anyone suggest me any other method to improve my frame rate preferentially to about 15 fps.
kindly give me your suggestion with an expected frame rate. would help me big time
Thanks in advance.
Regards
Abul Hassan
Hi Mohammed,
This solution seems to be ad-hoc for your topic.
http://www.mathworks.com/matlabcentral/fileexchange/46674-real-time-face-recognition-and-detection-system
Bye,
Alessandro
Hi Mohamed,
are you running the Viola-Jones detection in every frame? If yes, consider using the face detection only as an initialization for some sort of (feature-point) tracking algorithm - running a face detector in every single frame is computationally expensive. Tracking only a few features across frames will give you much faster results.
I found an example of Viola-Jones + KLT tracker with MATLAB, you can have a look in the link below. Although it uses the Computer Vision System Toolbox, which you might not have installed, it should give you an idea of the type of algorithm pipeline I explained.
http://de.mathworks.com/help/vision/examples/face-detection-and-tracking-using-the-klt-algorithm.html
Try to implement with V-Jones combination with KLT and apply detection method for a group of frames instead of single frame.
face detection using color information is the fastest approach, then if it is a possible face you may confirm using other more accurate methods such as the Viola-Jones.
and as Michael has said you do not have to check for every frame, do face tracking in the next frames, it is much faster.
Hi,
Check this site
http://www.face-rec.org/algorithms/
Regards
If your video stream is acquired in RGB color, you can consider reducing the false negative instances by working on a different color space and limit the search to those regions that match human skin chroma. This was also hinted by Ahmad Hassanat in the above post.
Option 2:
MATLAB is an interpreted language. However, it provides a bridge to low-level languages (C, C++) through its "mex" file generator which can highly likely accelerate the execution of the Viola-Jones algorithm.
Better Algorithm is from my perspective is:
Voila Jones "Integral Image based MB-LBP".
its faster in context of retaining more robust features due to Local and Global feature description.
To speed-up the face detection process:
First you can use motion and color analysis.
Second, for gray scale images, before applying your preferred face detection algorithm, you can use a simplest and faster face detection algorithm, which has an high detection rate and a high false alarm rate. The main face detection algorithm, which has a low false alarm rate is used to verify the proposed face images.
A last and efficient way to speed-up the face detection process is to use a grid on the image. During a first phase only the intersection points of the grid are tested. When the score or probability of a tested point is higher than a threshold a local and exhaustive search is done on the four adjacent squares. For instance if a 3x3 grid is used, the speed-up factor is around 9.
For more details see:
https://www.researchgate.net/publication/3193210_A_Fast_and_Accurate_Face_Detector_Based_on_Neural_Networks
Article A Fast and Accurate Face Detector Based on Neural Networks
@Raphaël Feraud
Thank you for your suggestion.
is it possible to get the MATLAB implementation of this paper
Would help me a lot
Thankyou
Unfortunately, there is no MATLAB implementation of this paper.
Regards,
Raphaël
@Raphaël Feraud
Dear,
I am having difficulty on implementing your paper, is there any kind of implementation I can use for the moment (any platform),
- Badges
- Science topic