Face Detector

A Simple Flask and docker-ready application that works as a server to detect faces, genders and their landmarks.

Docker

$ docker pull ahmdrz/facedetector:latest
$ docker run -p 5000:5000 ahmdrz/facedetector:latest

Sample

Request

$ curl "localhost:5000/detect?gender=on&landmarks=on&url=https://pixel.nymag.com/imgs/daily/vulture/2018/09/04/04-eminem-2.w700.h700.jpg

Response

{
  "result": [
    {
      "box": [
        0.3375,
        0.115,
        0.2275,
        0.2275
      ],
      "gender": "male",
      "index": 0,
      "landmarks": [
        [
          0.36,
          0.1875
        ],
        [
          0.3625,
          0.21
        ],
        [
          0.365,
          0.2325
        ]        
      ],
      "score": 0.43
    }
  ],
  "status": "ok"
}

NOTE Landmarks are 68 points.

Dependencies

First of all , clone this repository using git clone https://github.com/ahmdrz/facedetector.

To install all of dependencies run sudo pip install -r requirements.txt , after installing python packages , you have to download pre-trained dlib models.

  cd face_detection/data
  ./models.sh

How to use in clients ?

See example.py for details.

This small application has only one route , /detect.

Optional query parameters is landmarks and gender. Use landmarks=on to detect 68 point of face landmarks. Use gender=on to predict face gender.

You can use url query parameter.

Curl examples :

$  curl "localhost:5000/detect?url=<picture url>"
$  curl -F "image=@<picture file path>" "localhost:5000/detect?landmarks=on"
$  curl -F "image=@<picture file path>" "localhost:5000/detect?gender=on"

Note: All of the cordinates in this application is based of image width and image height.

Keywords

Face/Gender/Landmarks dectection
Docker face detection
Docker face gender detector
Docker face landmarks