faster_rcnn_face

Application of faster_rcnn in face recognition case

Setup by Franck FOTSO

Introduction:

This project aims to use the py_faster_rcnn tool to recognize face in real scene image. If your are interessed by the original project (python), please feel free to have a look here. We have setup a small dataset with only 24 classes, and launch a training with the VGG16 architecture. This repository involves all required files to launch a training or test process by your own. Indeed this page does not present the best model performance on our dataset, it just shows one more use case of the py_faster_rcnn project.

Datasets:

We have setup our own dataset (24 cls) based on pascal_voc schema. Required files are available here:

• ROMYNY Person 2017 : images+sets+annotations
• ROMYNY Person 2017 : labels

Hardwares/Softwares:

OS: Ubuntu 16.04 64 bit
GPU: Nvidia GTX 950M 4G
Cuda 8.0
CuDNN 3.0.8
Python 2.7.12
OpenCV 3.1.0

Prerequisites:

1. Caffe prerequisites
2. Python's packages (requirements.txt)

Installation:

To install this project, please follow the steps below:

1. Install OpenCV

2. Download the repository:

   $ cd /opt
   $ sudo git clone --recursive https://github.com/romyny/faster_rcnn_face.git
   $ cd faster_rcnn_face
   

3. Install all externals repositories required:

   $ cd caffe-faster-rcnn
   

   Adjust Makefile.config, then

   $ sudo mkdir build
   $ cd build
   $ cmake ..
   $ make -j4
   $ make install
   $ cd ../..
   

   Install project's libs

   $ cd libs
   $ make
   $ cd ..
   

4. Install python's packages required:

   for req in $(cat caffe-faster-rcnn/requirements.txt); do pip install $req; done
   

Get the data and models required:

1. Download the data and uncompress in 'data'

• person.24cls: Google Drive -> data/VOCdevkit2007

2. Download the models and uncompress in 'data/person_models'

• person_models (VGG16): Google Drive -> data/person_models
• person_models (RestNet-101): coming soon

Experiments:

1. Run the demo:

   $ cd $faster_rcnn_face_DIR
   $ python tools/demo.py 
   

   Note: this script will be performed on the test set

   For help use the command: python tools/demo.py --help

2. Launch train & test process:

   $ cd $faster_rcnn_face_DIR/experiments/scripts
   $ bash faster_rcnn_end2end_face.sh 0 VGG16 pascal_voc \
     --set EXP_DIR person_24cls RNG_SEED 42 TRAIN.SCALES "[400,500,600,700]"
   

Our results

The model generated at the 30k^th iteration gives a mAP^0.5 = 0.7217

All results related to demo dataset will be saved at 'data/demo_out'

Contact

Please feel free to leave suggestions or comments to Franck FOTSO (romyny9096@gmail.com)