FaceDatasetTools

Currently the GUI only supports Windows operating systems

Dependencies*(required)

• pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117
• pip install insightface==0.2.0
• onnx
• onnxruntime==1.15
• tqdm
• pillow
• numpy
• kmeans_pytorch
• pip install opencv-python
• pip install scikit-learn
• torchvision

Download*(required)

Download arcface_checkpoint.tar from: baidu cloud

Put arcface_checkpoint.tar to /arcface_ckpt

Download insightface_func.zip from: baidu cloud

Unzip insightface_func.zip to root path of this project (i.e., "./")

Usage

To crop faces from a video:

python face_crop_video.py

or

double click face_crop_video.bat

Parameters in GUI:

• Video Path: The path to the video file. Users need to create a dedicated folder for the target video to prevent the results from being overwritten.
• Target Path: Folder to store processed face images.
• Crop Size: Face image size.
• Align Mode: The padding size of the face area.
• Blurry Thredhold: The threshold for a picture to be considered blurry. The larger the value, the clearer the face.

To roughly classify different faces:

python face_cluster.py

or

double click face_cluster.bat

Detect face mask with FaRL:

Download FaRL.rar from: baidu cloud (password: we0q)