deepfacelive

DeepFaceLive

Input

• Source image

  

(Image from https://github.com/iperov/DeepFaceLive/tree/master/build/animatables)

• Driving image

(Image from https://github.com/iperov/DeepFaceLive/tree/master/build/samples)

Output

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 deepfacelive.py

If you want to specify the driving images, put the image directory path after the --input option.
The source image can be specified with the --source option.
You can use --savepath option to change the name of the output directory to save.

$ python3 deepfacelive.py --input DRIVING_IMAGE_DIR --source SOURCE_IMAGE --savepath SAVE_IMAGE_DIR

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.

$ python3 deepfacelive.py --video VIDEO_PATH

By adding the --detector option, you can specify detector type which is selected from "yolov5", "centerface", "s3fd". (default is yolov5)

$ python3 deepfacelive.py --detector yolov5

By adding the --marker option, you can specify marker type which is selected from "facemesh", "insightface". (default is facemesh)

$ python3 deepfacelive.py --marker facemesh

Reference

• DeepFaceLive

Framework

Onnxruntime

Model Format

ONNX opset=13

Netron

generator.onnx.prototxt