DeepFrontal: Deep learning-based 3D face reconstruction from 2D views

Le résumé fourni par la source

Construction of a realistic 3D frontal face from one 2D image is one of the fundamental problems in computer vision with usage in VR, animation, and biometrics. In this paper, DeepFrontal model is reviewed, which is a deep learning model that is an amalgamation of Generative Adversarial Networks (GANs) and Graph Convolutional Attention Networks (GCANs) to tackle the issues of pose variation, occlusion, and depth loss. Facial landmarks, a competent encoder, and mesh refinement for geometric and textural precision are incorporated into the framework. Tested on 300W-LP and AFLW2000-3D, DeepFrontal beats current state of the art in Chamfer distance and Mean Per-vertex Error, with photorealistic outputs. The model supports real-time deployment, pushing 3D face reconstruction to be more practical.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.