NormFace: L2 HyperSphere Embedding for Face Verification
https://www.cs.jhu.edu/~alanlab/Pubs17/wang2017normface.pdf
Announcement: Current arXiv version refers to a wrong result for A-softmax. I made a mistake when counting the depth of my CNN. The value 99.26% is for ResNet-20, while my CNN is ResNet-27.
| Baseline Model | Original Accuracy | Finetune Using Normalization |
|---|---|---|
| Light CNN(MaxOut) | 98.41%* | 98.78% |
| Center Face(ResNet) | 99.03% | 99.21% |
* It is 98.13% on Light CNN's project page. After applying the mirror face trick, it becomes 98.41%.
- My Caffe (https://github.com/happynear/caffe-windows/tree/ms). It also works on Linux. Anyway, if you want to use your own Caffe, please transplant the
inner_product_layer,inner_distance_layer,normalize_layer,general_contrastive_layer,flip_layerandaccuracy_layerto your Caffe. Since there are too many layers to transplant, I suggest you to use my Caffe directly:) - Matlab to draw some figures.
- GPU with CUDA support.
- MTCNN face and facial landmark detector(https://github.com/kpzhang93/MTCNN_face_detection_alignment).
- Baseline model such as Center Face or Light CNN or your own model trained by softmax loss. I don't know whether my loss helps or not if your model is trained by other loss functions. You may have a try. I believe it will be effective if there is no normalization term in your model.
- The dataset used in this paper is CASIA-Webface. Note that there are 3 identities overlap between CASIA-Webface and LFW. They are
0166921,1056413and1193098. For fair evaluation, it is recommended to remove them from CAISA-Webface. - Align all face images using MTCNN. The script can be found in my another github repository.
- Replace the final inner-product layer and softmax layer with layers defined in scaled_cosine_softmax.prototxt or normalized_Euclidean_contrastive.prototxt.
- Fine-tune the network based on the original model using a small learning rate, say 0.001 or 0.0001.
Evaluation codes are in my another github repository. Please refer to the second paragraph of the Update section.
A trick called mirror face is used during extracting the features. A sample code is in ./prototxt/example_of_mirror_face.prototxt.
About the mirror face: See ./MirrorFace.md
About the histogram feature for video face verification: Details are in ./YTF.
Light CNN B model(98.78%): Google Drive or Baidu Yun.
ps: I used to get a model of 98.8%, but I lost it....
Center Face (ResNet-27) model(99.21%): Google Drive or Baidu Yun.
This code is distributed under MIT LICENSE. The released models are only allowed for non-commercial use.
If you find my paper or the codes useful in your research, please cite
@article{Wang2017NormFace,
Author = {Feng Wang and Xiang Xiang and Jian Cheng and Alan L. Yuille},
Title = {NormFace: L2 Hypersphere Embedding for Face Verification},
Year = {2017},
Eprint = {arXiv:1704.06369}
}
Feng Wang [feng.wff(at)gmail.com], please replace (at) with @.