Skip to main content

Face Detection

Xircuits Face Detection Project Template

This template allows you to apply face detection on images or with webcam in real-time.

It consists of 2 components:

  • FaceDetector: Face detection with popular face detectors. It takes folder or image as input and detects face with speficied backend (default: mtcnn). If folder, result will be saved in output folder besides the same directory provided, else, result will be saved at the root directory here (Check the message printed out at the terminal).

  • CamInference: Camera inference with popular face detectors. It will open your webcam and detect face.

Prerequisites​

Python 3.9

If you would like to use dlib face detector:

  • Microsoft Visual Studio for C++
    (Note: Ignore the dlib installation error during installation if you want to skip this step. Also, avoid backend=2 or backend='dlib'.)

Installation​

  1. Clone this repository
git clone https://github.com/XpressAI/template-face_detection.git
  1. Create a virtual environment and install the required python packages
pip install -r requirements.txt
  1. Run xircuits from the root directory
xircuits

Workflow in this Template​

Inference​

FaceDetector.xircuits​

Where to Provide Input?

Input is taken by img_path of Component FaceDetector. Check out the example below.

You may provide folder or image file as input. Relative path or absolute path both are accepted (Noted that your folder in path should be separated by /).

Which Backend Should I Use?

RetinaFace and MTCNN seems to overperform in detection but they are much slower. If the speed of your pipeline is more important, then you should use opencv, ssd or mediapipe. On the other hand, if you consider the accuracy, then you should use retinaface, mtcnn or dlib.

Recommended for inference: retinaface, mtcnn or dlib.

Available backend:

  • 0 - opencv
  • 1 - ssd
  • 2 - dlib
  • 3 - mtcnn
  • 4 - retinaface
  • 5 - mediapipe
Where to Find Output?

If your input is folder, the output folder will be located besides your input folder. \ Example:

  • Input: resource/sample
  • Output: resource/output/sample

If your input is image file, the output will be located in this github root directory (the same directory as this README) at output.jpg and output.txt.

FaceDetector Example​

inference

Real-time Inference​

CamInference.xircuits​

Which Backend Should I Use?

Recommended for real-time inference: opencv, ssd or mediapipe.

Available backend:

  • 0 - opencv
  • 1 - ssd
  • 2 - dlib
  • 3 - mtcnn
  • 4 - retinaface
  • 5 - mediapipe
CamInference Example​

cam_inference

Possible Issue​

  1. Error to load weights while using 'ssd' backend
    Details: FileNotFoundError: [WinError 3] The system cannot find the path specified: 'C:\\Users\\user_name/.deepface/weights'
    Solution: Restart Xircuits.