radenkovic / cleverimage

MTCNN Face detection REST API [WIP]

cleverimage

THIS PROJECT IS STILL WIP!!!

Simple REST server for extracting features from images and suggesting tags.

Install

• create venv
• pip install requirements.txt
• $ source venv/Scripts/activate
• py server.js (runs bottle server)

Environment variables

Environment configuration:

Variable	Description	Required	Default value
ENV	If set to production server will run on gunicorn	no	N/A
PORT	Set up the server port	no	3000
FACE_DETECT_CONFIDENCE	Set up the confidence threshold for face_detect	no	0.8

API

Sample reqests are included in insomnia-data.json (for Insomnia client).

POST /face-detect

Detects faces in uploaded image.

{
  "endpoint": "/face-detect",
  "success": true,
  "data": [
    {
      "box": [ 3033, 1167, 431, 575 ],
      "confidence": 0.9996196031570435,
      "keypoints": {
        "left_eye": [ 3089, 1369 ],
        "right_eye": [ 3283, 1361 ],
        "nose": [ 3135, 1514 ],
        "mouth_left": [ 3124, 1612 ],
        "mouth_right": [ 3275, 1610 ]
      },
      "coordinates": {
        "x": [ 3033, 3464 ],
        "y": [ 1167, 1742 ]
      }
    }
  ]
}

TODO

• Cleanup the requirements (remove yolo model)
• Bottle image upload (multipart)
• Resize uploaded image (test for performance)
• Allow sending image URL