/Face-Recognition

Face recognition using TensorFlow

Primary LanguagePythonMIT LicenseMIT

Face-Recognition

This repository is a face recognition platform with TensorFlow.

It use CASIA as training set and LFW as test set.

Docker is recommended.

Code Structure

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├── outdated/           // Some old code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// Tool/outdated files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

LICENSE

MIT LICENSE

Author

irmowan

Date

2017.5