You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
git clone --recursive https://github.com/xmba15/face_recognition_survey
bash ./sh/preparation.sh
pip install -r requirements --user
# if you are using another version of cuda other than 10.0, try to install the right version of mxnet by fixing the related lines in requirements.txt
python ./scripts/extract_face_feature_from_image.py --image_path <image absolute path> --face_name <new name of the new face># (paramsは指定されない場合、デフォルトの値は使われます)
# csv result format would be# frame_num, object_1, position, [x1], [x2], [y1], [y2], pose, [yaw], [pitch], [roll], [top1_name], [dist1], [top2_name], [dist2], object_2,...# Only human objects that are recognized (which means they are detected and their poses and euclidean distances of two features vectors are within range of thresholds) are recorded in result file.
Test Accuracy on Dataset
# a small dataset to test accuracy is provided at images/test.# Run:
python scripts/test_accuracy.py
# to see the accuracy
Notes
# A notebook is also provided at scripts/test_accuracy.ipynb to test accuracy step by step
detector_wrapper
This system uses two models of face detector. SFDDetector and MtcnnDetector
SFDDetector is used fordetecting facein this system due to its speed and accuracy.
MtcnnDetector is used for getting the face landmarks