FaceSwap Implementation using PRNet

Worcester Polytechnic Institute | 2022

Overview

Course project implementing FaceSwap deep learning algorithms using Position Map Regression Network (PRNet) for 3D face reconstruction and alignment.

Description — 3D Head Mesh using normal map preview when faceswap

Note: Most images are not shown due to privacy concerns.

Before — Left: Face Mesh Mask with Depth based on video input. Right: Textureless Face Mesh

After — Left: Face Mesh Mask with Depth based on video input. Right: Textureless Face Mesh

Technical Details

Implemented FaceSwap using deep neural networks for face detection and replacement using PRNet
Collected custom dataset of faces and used it for training the FaceSwap model
Built FaceSwap script from scratch using PyTorch

© 2026 Jack (Haoying) Zhou. Website built and customized by Jack (Haoying) Zhou from the academicpages template. Please let me know if you notice any glitches.