• 设为首页
  • 点击收藏
  • 手机版
    手机扫一扫访问
    迪恩网络手机版
  • 关注官方公众号
    微信扫一扫关注
    迪恩网络公众号

dinhquy94/face-antispoofing-using-mobileNet: Face anti-spoofing using MobileNet

原作者: [db:作者] 来自: 网络 收藏 邀请

开源软件名称(OpenSource Name):

dinhquy94/face-antispoofing-using-mobileNet

开源软件地址(OpenSource Url):

https://github.com/dinhquy94/face-antispoofing-using-mobileNet

开源编程语言(OpenSource Language):

Python 100.0%

开源软件介绍(OpenSource Introduction):

CNN for face-anti-spoofing

The face anti-spoofing is an technique that could prevent face-spoofing attack. For example, an intruder might use a photo of the legal user to "deceive" the face recognition system. Thus it is important to use the face anti-spoofint technique to enhance the security of the system.
facial-anti-spoofing

All the face images listed below are in the dataset of CASIA-FASD

Why MobileNet?

MobileNet is an architecture which is more suitable for mobile and embedded based vision applications where there is lack of compute power. This architecture was proposed by Google.

Usage

Main Dependencies

Python 3 and above
tensorflow 1.3.0
numpy 1.13.1
tqdm 4.15.0
easydict 1.7
matplotlib 2.0.2
pillow 5.0.0

Prepare Dataset

CASIA-FASD datasets are consist of videos, each of which is made of 100 to 200 video frames. For each video, I captured 30 frames (with the same interval between each frame). Then, with the Haar_classifier, I was able to crop a person’s face from an image. These images make up training datasets and test datasets.

Train and Test

  1. Prepare your data, and modify the data_loader.py/DataLoader/load_data() method.
  2. Modify the config/test.json to meet your needs.

Note: If you want to test that the model is pretrained and working properly, I've added some test images from different classes in directory 'data/'. All of them are classified correctly.

Run

python3 main.py --config config/test.json

The file 'test.json' is just an example of a file. If you run it as is, it will test the model against the images in directory 'data/test_images'. You can create your own configuration file for training/testing.

Benchmarking

In my implementation, I have achieved approximately 1140 MFLOPS. The paper counts multiplication+addition as one unit. My result verifies the paper as roughly dividing 1140 by 2 is equal to 569 unit.

To calculate the FLOPs in TensorFlow, make sure to set the batch size equal to 1, and execute the following line when the model is loaded into memory.

tf.profiler.profile(
        tf.get_default_graph(),
        options=tf.profiler.ProfileOptionBuilder.float_operation(), cmd='scope')

I've already implemented this function. It's called calculate_flops() in utils.py. Use it directly if you want.

Updates

  • Inference and training are working properly.

Test

test




鲜花

握手

雷人

路过

鸡蛋
该文章已有0人参与评论

请发表评论

全部评论

专题导读
热门推荐
阅读排行榜

扫描微信二维码

查看手机版网站

随时了解更新最新资讯

139-2527-9053

在线客服(服务时间 9:00~18:00)

在线QQ客服
地址:深圳市南山区西丽大学城创智工业园
电邮:jeky_zhao#qq.com
移动电话:139-2527-9053

Powered by 互联科技 X3.4© 2001-2213 极客世界.|Sitemap