Image & Vision

Image & Vision

Investigates Image data fusion techniques and video analytics that combine image and track data from multiple sensors to achieve improved accuracies and more specific inferences than could be achieved by using a single sensor alone. Our aim is to explore the state-of-the-art image processing and video analytics algorithms for achieving effective enhancement, detection, tracking, and video summarization as in: 

3D Face Reconstruction

1. Introduction

• 3D face reconstruction by Generative Adversarial Networks (GANs) has been remarkably successful. However, there are still some challenges in training data acquisition, high-fidelity UV map generation, other styles of UV map generation. In particular, there is a lack of research on generating diverse styles of UV maps, such as toonify and Disney styles. we leverage synthesized multi-view images and their predicted 3D information to generate texture-rich and high-resolution UV facial textures from a single portrait. Moreover, we can generate different styles of UV maps by plugging ’Map and edit’ into other fine-tuning StyleGAN2.

2. Research

• We can accomplish the following: (1) Generate a 3D face mesh from a single 2D photo; (2) Generate a variety of style3D face meshes from a single 2D photo; (3) 3D facial mesh animation using speech audio; (4) Synthesize full 3D face mesh from a single 2D photo.