Face Extraction And Description

This simple app aims to extract faces from the stream (live camera stream or existing video file or even the bunch of frames etc.) and shows you all possible information about the person accordingly to his facial features.

Currently supports age and gender recognition.

Requirements

• Ubuntu (tested 16.04)
• Python 2.x (tested 2.7.12)
• OpenCV 2.4.x (tested 2.4.12.3, manual build)
• Caffe with Python bindings
• Dlib with Python bindings
• NumPy
• SciPy
• Click

Installation

1. Install OpenCV from the repository (the easiest way) or as described here.

2. Build Caffe manually as described here. Don't forget about Python bindings:

   make pycaffe

3. Run

   sudo pip install -U -r requirements.txt

   from the root of the repository.

HowTo

1. Camera calibration

   • fill the section calibration in the your config file (see configs/config.yml.example for possible parameters) with the appropriate values

   • run

     ./calibrate.sh PATH_TO_CONFIG

   Calibrations will be saved in the calibrations subfolder with name specified in the config.

2. Downloading pre-trained models.

   Currently this solution supports two models both for age and gender recognition from two different papers:

   • Age and Gender Classification using Convolutional Neural Networks. Gil Levi, Tal Hassner, 2015 [1]

     This model trained on Adience dataset.

   • DEX: Deep EXpectation of apparent age from a single image. Rasmus Rothe, Radu Timofte, Luc Van Gool, 2015 [2]

     This model based on the pre-trained one on ImageNet2014 dataset [3] and then fine-tuned on IMDB-WIKI. Actually this project uses the model which was additionally fine-tuned on the LAP challendge dataset.

   This repository already provides network declarations (see models subfolder), pre-trained weights can be donwloaded by running

   ./download_models.sh

3. Running the app

   • fill the section app in the your config file (see configs/config.yml.example for possible parameters) with the appropriate values

   • run

     CAFFE_ROOT=/path/to/caffe/ ./run.sh PATH_TO_CONFIG

   You should see a resizable window with your stream (camera or video file) overlaid by the facial b-boxes and facial descriptions.

References

1. Project page

2. Project page

   @article{Rothe-IJCV-2016,

   author = {Rasmus Rothe and Radu Timofte and Luc Van Gool},

   title = {Deep expectation of real and apparent age from a single image without facial landmarks},

   journal = {International Journal of Computer Vision (IJCV)},

   year = {2016},

   month = {July},

   }

   @InProceedings{Rothe-ICCVW-2015,

   author = {Rasmus Rothe and Radu Timofte and Luc Van Gool},

   title = {DEX: Deep EXpectation of apparent age from a single image},

   booktitle = {IEEE International Conference on Computer Vision Workshops (ICCVW)},

   year = {2015},

   month = {December},

   }

3. Project page

   Very deep convolutional networks for large-scale image recognition. K. Simonyan and A. Zisserman, 2014.