• The origin mobilenetV2 code is here mxnet-mobilenet

• created by : lxy and shj
• Time: 2018/12/10 15:09
• project Face Anti Spoofing
• company:
• rversion: 0.1
• tools: python 2.7
• modified:
• description: The codes for training and testing

• mxnet >= 1.2.0
• python >= 2.7.15
• opencv >= 3.4.0

• The training datas are downloaded from internet,using the tool BaiduDownload
• We have created the dataset including 4 classes (Mobilephone:1 TV:2 telectroller:3 background:0).

Configuration parameters lies in Root/src/configs/config.py

1. directory

• data is used to store training and testing data.
• log is used to store traing logs.
• models is used to store network parameters.
• src is used to store training and testing codes.

2. train

• get image list : running Root/src/utils/run.sh to generate traing and testing data list.
• pack training images: running Root/src/prepare_data/convert2data.sh to pack training data.
• to train on packed images: running Root/src/train/train_faceanti.py

3. test

• test one image: python Root/src/test/demo.py --img-path1 test.jpg --gpu 0 --load-epoch 10 --cmd-type imgtest
• test a video: python Root/src/test/demo.py --file-in test.mp4 --gpu 0 --load-epoch 10 --cmd-type videotest
• test on a test dataset: python demo.py --file-in ../../data/test.lst --out-file ./output/record.txt --base-dir .../test_imgs/ --load-epoch 25 --cmd-type txtlisttest