Face detection is to locate faces in a image. It can be seen as pre-processing of face recognition/verification/identification. Traditional methods include Haar-cascade method and HOG based method. I suppose that the state-of-art method currently is deep learning based method. Does anyone know which deep learning method is with the best performance? Thanks.
Dear Bo Li ,
Read my article, please.
https://www.researchgate.net/publication/268003599_INFORMATION_IDENTIFICATION_SYSTEM_FOR_IDENTIFYING_PEOPLE_BY_PORTRAIT_PHOTOS
https://www.researchgate.net/publication/220499255_The_effectiveness_increase_of_a_system_of_automatic_biometrical_identification_based_on_photo_portraits
Regards, Shafagat
Article INFORMATION IDENTIFICATION SYSTEM FOR IDENTIFYING PEOPLE BY ...
Article The effectiveness increase of a system of automatic biometri...
Your answer strongly depends on the purpose of your research. Convolutional Neural Networks works very well for qualitative research (emotions, facial structure, identification of specific characteristics, identification of people) but an Artificial Neural Network is a better option for quantitative analysis (valence and arousal). Also it depends on the validation method available for your face detection algorithm (standarized databeses, iMotions FACET platform, questionnares). A little bit about the type of research you do (identification of qualitative or quantitative data, validation method) could help us to give a better answer.
We recently performed a face detection competition: https://arxiv.org/abs/1708.02337
The winner of the competition is the TinyFaces algorithm (which is based on Deep Neural Networks), available online here: https://www.cs.cmu.edu/~peiyunh/tiny
BTW: I have seen many people calling the Viola-Jones detection based on Harr features. This is a spelling mistake, it should be Haar features.
Viola Jones Algo - compatible with computer vision tool box in the matlab
@Smiriti Tikoo and @A.Al-Naji: The inquirer specifically asked for state-of-the-art neural-network-based face detection. Viola Jones is neither state of the art, nor is it DNN-based. He also did not mention that he uses Matlab, so being compatible with Matlab is not required. Please read the question before you answer!
@Manuel Günther
I'm looking for a good face detector,which I can run from python (the best performing algorithm runs on Matlab),preferably by using Pytorch (or TF), and performs better than dlib (and opencv) detectors, which I am currently using.
Any good suggestions for that?
Manuel Günther
I'm looking for a good face detector,which I can run from python (the best performing algorithm runs on Matlab),preferably by using Pytorch (or TF), and performs better than dlib (and opencv) detectors, which I am currently using.
Any good suggestions for that?
@Manuel Günther
I'm looking for a good face detector,which I can run from python preferably by using Pytorch (or TF), and performs better than dlib (and opencv) detectors, which I am currently using.
Any good suggestions for that?
@Manuel Günther
I'm looking for a good face detector,which I can run from python,preferably by using Pytorch if any framework is needed and performs better than dlib and opencv detectors, which I am currently using.
Any good suggestions for that?
Yoni Keren
Sorry, I just read your reply. There is the MTCNN face detector, which is originally implemented in Matlab, but where python wrappers (using Caffe though) exist: https://github.com/walkoncross/mtcnn-caffe-zyf
This is one of the state-of-the-art face detectors currently available.
I hope this helps
Manuel
Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Networks is considered state-of-the-art face detection.
I have used this detector in my two works:
Git Loss for Deep Face Recognition: Preprint Git Loss for Deep Face Recognition
A Pipeline to Improve Face Recognition Datasets and Applications: https://www.researchgate.net/publication/328382309_A_Pipeline_to_Improve_Face_Recognition_Datasets_and_Applications
Some other face recognition work that use this face detector: https://ydwen.github.io/papers/WenECCV16.pdf
- Badges
- Science topic
- Similar topics
- Analytical Chemistry
- Extraction