Face-Detection

Face detection based on haar, dlib, and mtcnn detectors

Introduction

A simple implementation of face detection (Dlib/OpenCV/MTCNN backend)

---

Quick Start

1. Clone this repo.
2. Run face detection.

git clone https://github.com/Ianpanda/Face-Detection.git
python face_detection.py -i demo_images/00.jpg for image detection mode, OR
python face_detection.py -i [video_path] -o [output_path (optional)] for video detection mode.

Usage

Use -h or --help to see usage of face_detection.py:

usage: face_detection.py [-h] [-i INPUT] [-o OUTPUT]
                         [-d {haar,dlib,mtcnn,all}] [-s SPEED] [-v {y,n}]

detect faces based on haar or dlib

optional arguments:
  -h, --help            show this help message and exit
  -i INPUT, --input INPUT
                        video or image to detect
  -o OUTPUT, --output OUTPUT
                        video or image result
  -d {haar,dlib,mtcnn,all}, --detector {haar,dlib,mtcnn,all}
                        choose your detector
  -s SPEED, --speed SPEED
                        spped up video result if need
  -v {y,n}, --visualize {y,n}
                        visualize the detect result immediately

demo：python face_detection.py -i file_in -o file_out -d detector
1.-i:video or image to detect
2.-o:result
3.-d:choose your detector
4.-s:spped up video result if need
5.-v:visualize the detect result immediately

---

Some issues to know

1. The test environment is

   • Python 3.5.4
   • OpenCV 3.3.1
   • Dlib 19.15.0
   • TensorFlow 1.9.0

2. Default haar classifier are used. If you use others, try to get from $ROOT/model_weights/haar, or you can train your own xml.

3. The MTCNN model has fixed weights in model_weights/mtcnn.

---

To Do List

• Update method to train xml
• Add MTCNN detector
• Extend other amazing Networks