face-anti-spoofing

Lightweight Face Anti Spoofing

Input

(Image from https://search.creativecommons.org/photos/df3a19c2-47ca-4f58-8aed-0dc62e89e9e9)

Shape: (1, 3, 128, 128) RGB channel order

Output

• Estimating spoofing

### Estimating spoofing ###
face is real: 98.432%

• Detected face estimation

Usage

Automatically downloads the onnx and prototxt files on the first run. It is necessary to be connected to the Internet while downloading.

For the sample image,

$ python3 face-anti-spoofing.py 

If you want to specify the input image, put the image path after the --input option.

$ python3 face-anti-spoofing.py --input IMAGE_PATH

If you want to perform face detection in preprocessing, use the --detection option.

$ python3 face-anti-spoofing.py --input IMAGE_PATH --detection

By adding the --video option, you can input the video.
If you pass 0 as an argument to VIDEO_PATH, you can use the webcam input instead of the video file.
You can use --savepath option to specify the output file to save.

$ python3 face-anti-spoofing.py --video VIDEO_PATH --savepath SAVE_VIDEO_PATH

Reference

• Lightweight Face Anti Spoofing

Framework

Pytorch

Model Format

ONNX opset = 11

Netron

MN3_large.onnx.prototxt