文件列表:

data (0, 2018-09-26)
data\pairs.txt (155335, 2018-09-26)
dataset (0, 2018-09-26)
dataset\cleaned_list.txt (9942040, 2018-09-26)
lib (0, 2018-09-26)
lib\lfw.py (3159, 2018-09-26)
lib\utils.py (30017, 2018-09-26)
lr_coco.txt (124, 2018-09-26)
lr_scatter.txt (116, 2018-09-26)
networks (0, 2018-09-26)
networks\inception_resnet_v1.py (11810, 2018-09-26)
networks\resface.py (5181, 2018-09-26)
networks\sphere_network.py (3852, 2018-09-26)
test.sh (609, 2018-09-26)
test (0, 2018-09-26)
test\test.py (8471, 2018-09-26)
train.sh (3042, 2018-09-26)
train (0, 2018-09-26)
train\train_multi_gpu.py (23063, 2018-09-26)

## Recent Update ```2018.07.04```: I achieved a better accuracy(99.2%,[trained model](https://pan.baidu.com/s/1c7bPoM_hGvkzp5Tunu_ivg)) on LFW. I did some modification as bellow: - Align webface and lfw dataset to ```112x112```([casia-112x112](https://pan.baidu.com/s/1MYNq6pkZJCkpKERC92Ea1A),[lfw-112x112](https://pan.baidu.com/s/1-QASgnuL0FYBpzq3K79Vmw)) using [insightface align method](https://github.com/deepinsight/insightface/blob/master/src/align/align_lfw.py) - Set a bigger margin parameter (```0.35```) and a higher feature embedding demension (```1024```) - Use the clean dataset and the details can be seen [this](https://github.com/happynear/FaceVerification/issues/30) ## CosFace This project is aimmed at implementing the CosFace described by the paper [**CosFace: Large Margin Cosine Loss for Deep Face Recognition**](https://arxiv.org/pdf/1801.09414.pdf). The code can be trained on [CASIA-Webface](http://www.cbsr.ia.ac.cn/english/CASIA-WebFace-Database.html) and the best accuracy [LFW](http://vis-www.cs.umass.edu/lfw/) is ***.6%. The result is lower than reported by paper(99.33%), which may be caused by sphere network implemented in tensorflow. I train the sphere network implemented in tensorflow using the softmax loss and just obtain the accuracy 95.6%, which is more lower than caffe version(97.88%). ## Preprocessing I supply the preprocessed dataset in baidu pan:[CASIA-WebFace-112X96](https://pan.baidu.com/s/160RN84j_79TnktKZmzakfw),[lfw-112X96](https://pan.baidu.com/s/1fkH9xR5Z0inxTP7Maae2KQ). You can download and unzip them to dir ```dataset```. If you want to preprocess the dataset by yourself, you can refer to [sphereface](https://github.com/wy1iu/sphereface/tree/0056a7d27d05f2815a276cb2***71f0348d6dd8da#installation). ## Train ```./train.sh``` ## Test Modify the ```MODEL_DIR``` in ```test.sh``` and run ```./test.sh```. If you do not want to train your model, you can download my [trained model](https://pan.baidu.com/s/1ouQA2PXz1hp7Uz_uhsyMdw) and unzip it to ```models``` dir. ## Reference - [facenet](https://github.com/davidsandberg/facenet)

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：