Sardar Vallabhbhai National Institute of Technology
Question
Asked 11th Nov, 2016
Which is the best algorithm for Face Recognition?
PCA algo
Fishersface algo
HMM Which is best in terms of computation as well as accuracy.
Most recent answer
ArcFace is the best face recognition model in the following scenario:
1) Constraint and Unconstraint
2) Low resolution, Blurry, Pose Invariant, illumination
3) Age-Invariant or Cross-Age Face recognition (Adult, Child, Diverse Ethnicity)
4) Mask Wearing face recognition.
5) Aging or Longitudinal dataset
Check This paper for face verification and identification result on dataset
1 Recommendation
Popular answers (1)
Bells University of Technology, Ota, Ogun State
Dear respected Surjit Paul,
Other methods used for face recognition are Deep Hypersphere Embedding for face recognition, FaceMatch, EigenFaces, FischerFaces, LBPH, Face Everthing, Fisherfaces, Real Time Face Recognition, DeepID Test, Node FaceNet, Facial Recognition API for Python and Command Line, Tensorflow, Joint Face Detection and Alignment, OpenBR, OpenFace, Lambda Labs Face Recognition API, OpenFace, Google Vision API, SkyBiometry, Amazon Rekognition, MS Azure Face API, SphereFace, Deep Face Recognition with Caffe Implementation and Face Recognition in ARKit.
Best regards.
Kifilideen.
5 Recommendations
All Answers (26)
Coventry University
Surjit,
At the very early stage, you need to apply PCA method. After that, if your aim is selecting between either "Fishersface" or "HMM", then I suggest choosing the one produces the best prediction results.
HTH.
Samer
1 Recommendation
Washington University in St. Louis
For face recognition, you need to follow deep learning algorithms. In my opinion, deep learning artificial neural network is the best...
Islamic Azad University, nour branch
dear Surjit
PCA and fisherFace in particular test is not good (in ORL face database is good)
for face detection adaBoost(viola-jons article) is very good
University of Zurich
What you are asking is very difficult to answer. The answer highly depends on the type of images that you are dealing with, and what is the scale of your database. For example, when you have mainly frontal faces in your database, PCA and LDA might be good solutions, but others like Gabor graphs or LGBPHS should work, too. Implementations of these (in Python) can be found here: https://gitlab.idiap.ch/bob/bob.bio.face, which will also deal with the face detection and alignment, if necessary.
However, all of these algorithms require frontal faces. If you have uncontrolled data and non-frontal poses, you should consider using a deep neural network, for example as provided by the VGG group: http://www.robots.ox.ac.uk/~vgg/software/vgg_face or in OpenFace (see link of Mohan). I have not seen any other algorithm being able to recognize non-frontal faces.
2 Recommendations
Try haar classifier using opencv . Its good and quite accurate.
Its also very light so that you can apply it real time.
1 Recommendation
Jadavpur University
This is not easy to answer, as there is no one and only answer for this. People are using different features over the years. Outcomes depend on various parameters.
Anyway as others are saying Deep Learning is a common trend. You can try.
If the GPU is not available you can try some new features developed by the researchers. Like you can give a look (a recent paper):
1 Recommendation
Novosibirsk State University
Try to use OpenBR: http://openbiometrics.org/
It implements 4SF algo for face recognition and comparison with high accuracy from the box.
Bells University of Technology, Ota, Ogun State
Dear respected Surjit Paul,
There are different kinds of methods used for Face Recognition, but the best are based on Deep Learning algorithms. They are commonly used these days. The deep learning algorithms project a face image in a 128-dimensional euclidean space.
Best regards.
Kifilideen.
4 Recommendations
Bells University of Technology, Ota, Ogun State
Dear respected Surjit Paul,
Other methods used for face recognition are Deep Hypersphere Embedding for face recognition, FaceMatch, EigenFaces, FischerFaces, LBPH, Face Everthing, Fisherfaces, Real Time Face Recognition, DeepID Test, Node FaceNet, Facial Recognition API for Python and Command Line, Tensorflow, Joint Face Detection and Alignment, OpenBR, OpenFace, Lambda Labs Face Recognition API, OpenFace, Google Vision API, SkyBiometry, Amazon Rekognition, MS Azure Face API, SphereFace, Deep Face Recognition with Caffe Implementation and Face Recognition in ARKit.
Best regards.
Kifilideen.
5 Recommendations
TetherBox Technologies
Undoubtedly, there are many such software found in the field of face recognition. I can suggest the best for you depending on the amount you wish to spend and where you would implement the software.
For Now, the one I liked is FaceX (Face Recognition APIs for your apps). They have good support, and the pricing compared to the rest of the same class in the market.
You may give it a try.
Government College of Engineering Erode
Different Algorithms gives different accuracy.
Using Deep Learning, Machine Learning techniques: Convolutional Neural Network.
Use ANFIS.
Use BPN ANN, SOM ANN.
Use Genetic Algorithm with PCA.
Use statistical Appearance methods-EigenFace, FisherFace,2DPCA,ICA,KernelPCA, Bag of Visual Methods other SIFT based methods
20Face
FaceNet (https://github.com/davidsandberg/facenet) is best for face recognition as of Jan 2020.
1 Recommendation
Hanoi University of Industry, Vietnam
ARCFace is the best now. ( https://medium.com/@peaceful0907/face-recognition-arcface-cf814fb5549e )
1 Recommendation
20Face
Thanks Chi-Kien Tran .
Its strange, we see no improvement in the past two years.
Manipal University Jaipur
Hello,
Can anyone suggest which model can we use for finding lost person using cctv camera
RonGlad Environmental Engineering Services Co. (RGEES)
I am not quite familiar with the topic but these kinds of technologies have been applied to all establishments for security purposes. I think Sir Kifilideen Lekan Osanyinpeju had enumerated the system of the best algorithm for Face Recognition.
Thanks
VNU University of Science
In my experience, I think this question depends on a lot of context to get good answer , like : which purpose let you recognize face, all over the world or just a continent... Usually i usually use vector (extract feature techinque) from a DNN model like mxnet ,facenet and tuning more by your specific NN to adapt with your purpose!
Sardar Vallabhbhai National Institute of Technology
Best CNN based face recognition(Verification and Identification) matcher:
1) FaceNet
2) Probablisit Face Embedding
3) ArcFace
4) Cosface
5) Spherface
The scenario where we apply the face recognition CNN face matcher.
ChildFace: Gender Aware Child Face Aging
20Face
This one of the most accurate and open-source face recognition algorithm https://github.com/deepinsight/insightface
Sardar Vallabhbhai National Institute of Technology
Arcface is the best performance on low resolution, mask face, pose invariant, and cross-age face recognition.
ArcFace face verification example in the context of Cross-Age Face Recognition:
Conference Paper Child Face Age Progression and Regression using Self-Attenti...
20Face
Praveen Kumar Chandaliya which github repository do you suggest for using Arcface?
Sardar Vallabhbhai National Institute of Technology
ArcFace is the best face recognition model in the following scenario:
1) Constraint and Unconstraint
2) Low resolution, Blurry, Pose Invariant, illumination
3) Age-Invariant or Cross-Age Face recognition (Adult, Child, Diverse Ethnicity)
4) Mask Wearing face recognition.
5) Aging or Longitudinal dataset
Check This paper for face verification and identification result on dataset
1 Recommendation
Similar questions and discussions
Related Publications
A novel ambient intelligence interactive technique are introduced here. First, the mechanism of our AmI system and several idiographic technique issue is expounded stepwise. For providing personalized service an algorithm of face recognition based on HMM is researched and the experimental results show efficient. In this way the user can be identifi...
A transform domain face recognition approach is presented. The DCT is coupled with the HMM to achieve a recognition rate of 100% on ORL face database of 40 subjects with 10 images per subject. The recognition time for ORL database is little over 2 Sec. 5 images of a subject are used to train HMM and remaining 5 are used for recognition test. The pr...
The work presented in this paper focuses on the use of Hidden Markov Models for face recognition. A new method based on the extraction of 2D-DCT feature vectors is described, and the recognition results are compared with other face recognition approaches. The method introduced in this paper reduces significantly the computational complexity of prev...