This repository is a face recognition platform with TensorFlow.
It use CASIA as training set and LFW as test set.
Docker is recommended.
Here is the main code structure of the repository.
Attention: the code is under TensorFlow 1.1 and python 2.7.
// Directories
├── dataset/ // Datasets, including CASIA, LFW
├── tfrecord/ // TFRecord for tensorflow
├── images/ // Output images
├── outdated/ // Some old code
├── tools/ // Some tool files
├── train_data/ // Store train data
├── features/ // Store features when test
└── txt/ // Text files for data
// Main files
├── make_tfrecords.py // Make images into one tfrecord file
├── multi_gpu_train.py // Main training file for multi-gpu train
├── trans_config.py // Config file for transform
├── transform.py // Image transform using cv2
├── pca_fuse.py // Feature fuse and PCA
└── test_lfw.py // Test LFW using searching threshold
// Tool/outdated files
// You can find some ideas or here.
├── input.py
├── pca.py
├── vgg.py
├── fuse_feature.py
├── filter_file.py
└── filter_lfw.py
// Test files
// These files are for exporting images
├── test_align.py
└── test_tfrecord_image.py
2017.5