Face recognition

reference

https://github.com/openvinotoolkit/open_model_zoo/tree/master/demos/face_recognition_demo/python

depndency

pip install openvino-dev==2022.1.0

model

• Face detection: face-detection-adas-0001

  input: 1x3x384x672 BGR image

  output: 1x1xNx7, where N is the numver of detected face, 7 contain [image_id, label, confidence, x, y, w, h]

• Landmark detection: landmarks-regression-retail-0009

  input: 1x3x48x48 BGR image

  output: 1x10, (x0, y0, x1, y1, ..., x5, y5)

• Face reidentification: face-reidentification-retail-0095

  input: 1x3x128x128 BGR image

  output: 1x256x1x1, containing a row-vector of 256 floating point values

setup

git clone git clone https://github.com/openvinotoolkit/open_model_zoo.git
pip install .\demos\common\python\
cd <omz_dir>/demos/face_recognition_demo/python/
omz_downloader --list models.lst
omz_converter --list models.lst
mkdir face_gallery

prepare

place face image into face_gallery folder

python face_recognition_demo.py -i 0 -m_fd .\intel\face-detection-adas-0001\FP16-INT8\face-detection-adas-0001.xml -m_lm .\intel\landmarks-regression-retail-0009\FP16-INT8\landmarks-regression-retail-0009.xml -m_reid .\intel\face-reidentification-retail-0095\FP16-INT8\face-reidentification-retail-0095.xml --verbose -fg .\face_gallery\ --run_detector

enter name. press enter to save or esc to exit

Usage (python)

python face_recognition_demo.py -i 0 -m_fd .\intel\face-detection-adas-0001\FP16-INT8\face-detection-adas-0001.xml -m_lm .\intel\landmarks-regression-retail-0009\FP16-INT8\landmarks-regression-retail-0009.xml -m_reid .\intel\face-reidentification-retail-0095\FP16-INT8\face-reidentification-retail-0095.xml --verbose -fg .\face_gallery\ -d_fd GPU -d_lm GPU -d_reid GPU

Usage (c++)

1. place cropped face image into database folder
2. build solution and run