Generatation a composite baby face from two parent images

Description

This project presents a method to generate a composite baby face from two parent images.

Face detection, facial landmark identification, and feature extraction are first implemented to collect the relevant information from each parent image. 

Using this information, a facial morphing algorithm including the partitioning of features into triangles and quadrilaterals based on feature keypoints, warping triangles and quadrilaterals using affine transformations and homography respectively, and cross dissolving the warped images is implemented. The composite adult features are then combined with the base baby features and placed onto the base baby face at the corresponding locations.

Example results of generated baby images. The right column shows the generated baby image of the parent images in the left and middle columns

Feature SIMilarity (FSIM) index between two images