phanducdung888 / eye-pair Goto Github PK

1. Clone this repository.

2. Install requirements: pip install -r requirements.txt

3. If input face images are in directory example/input/dir and output direcotry example/output/dir then run the following command which will preserve input directory structure for further tasks.

   python crop_eye_pair.py --input_dir 'example/input/dir' --output_dir 'example/output/dir' --method 'mtcnn' --device 'cuda'
   

4. Supported method options: haar-cascade and mtcnn

5. Supported device options: cpu and cuda.

Notes: Setting device to cuda has no effect while using haar-cascade method.

1. MTCNN code borrowed from https://github.com/khrlimam/mtcnn-pytorch
2. HAAR Cascade code inspired from https://github.com/zekeriyafince/EyePair