Facedetection Experiments

This repo is for testing the basics of faceswap.

1. Facedetection

There are three different ways to detect the face:

• Haar Detector
• Histogram of Oriented Gradients
• Neural Network

Somehow, all seem to work good...

2. Facial Landmarks

The pose of the face is estimated by fitting 68 facial landmarks to significant points in the face. From those points the face can be divided in triangles.

3. Swap Faces

Each triangle is then warped to the position of the corresponding triangle in the other face and smoothed a bit.