This project is to automatically measure the facial Width-To-Height Ratio and get facial analysis results provided by Microsoft Azure Cognitive Services. Used in

• The face of risk: CEO facial masculinity and firm risk by Shinichi Kamiya, Y. Han (Andy) Kim, Soohyun Park (Me)
• "부자, 관상, 기술" written by Y. Han (Andy) Kim

You can get some estimatations of below (per image):

• Facial Width-To-Height Ratio

• X, y coordinates for 68 facial landmarks

• If you use Microsoft Face APIs:

  • Age
  • Gender
  • Emotions
    • fear, sadness, disgust, contempt, neutral, happiness, anger
  • Glasses / no glasses
  • Moustache
  • Beard
  • Sideburns
  • Bald

Since June 11th, 2020, Microsoft Azure Face recognition services are strictly prohibited for use by or for US police departments. Read this announcement

• Python 3.6
• Dlib
• Opencv
• Flask

1. python version 3.6 with Anaconda distribution (no guarantee for other versions)
   • You can download Anaconda Individual Edition in here
   • Check your anaconda installation by conda -V
   • Create a virtual environment by conda create -n [name] python=3.6 and activate the venv by conda activate [name]
2. Clone this repo
   • git clone https://github.com/haileypark-kr/fwhr-calc-website.git
   • Move to the project root directory. cd fwhr-calc-website
3. Microsoft Azure Face Api Key
   1. Create an Azure account and a Cognitive Service Face API resource in Azure Portal. Read this documentation.
   2. Generate keys to access your API. (Resource Management > Keys and Endpoint)
   3. Make a file named azure_faceapi_key.conf and paste the first key in the file. (you can change the file name if you want, but make sure you also change .gitignore and config.py) Do not upload this file to GitHub.
   4. Replace the variable FACE_API_ENDPOINT in config.py with your endpoint.

      # config.py
      
      FACE_API_ENDPOINT = "https://eastasia.api.cognitive.microsoft.com"
      

Install python libraries in this project's root directory.

1. pip install -r requirements.txt
2. conda install -y -c conda-forge dlib
   • Some libraries (e.g. dlib) cannot be installed by pip - should be installed using conda.

There are two ways to run this application.

• Running a flask web server: If you want to analyze a few facial images with GUI.
• Running fWHR calcaculating script: If you want to analyze thousands of images.

1. Command: python app.py
2. Open a Chrome browser and enter 127.0.0.1:5001
3. Select some images and press Submit button.
4. Wait and do not reload the browser.
5. Anlysis result will be downloaded shortly. (in xlsx format)

1. Command: python fWHR_main.py --dataroot [path to the image directory]
2. Wait
3. Go to data/output direcetory and get the analysis result file.

You can change configuration values configured in config.py.

1. directory path configs

   • UPLOAD_FOLDER: directory to save uploaded images via flask web server
   • OUTPUT_FOLDER: directory to save output images
   • ERROR_DATA_FOLDER: directory to save images where this application couldn't find any faces
   • PRETRAINED_MODEL_PATH: path to pretrained model. Either a relative path from this project root directory or an absolute path.

2. image configs

   • ALLOWED_EXTENSIONS: set of allowed image extensions. case insensitive.

3. Microsoft Azure configs

   • USE_MS_FACE_API: boolean value whether to use Microsoft Face API or not. (False / True)
   • FACE_API_KEY_CONFIG_FILE: name of the config file storing Microsoft Face API key. You can get this value in Azure portal.
   • FACE_API_ENDPOINT: endpoint of Microsoft Face API. You can get this value in Azure portal.
   • FACE_API_URL: full url for Microsoft Face API. Do not edit this value unless the API's url path is changed.