A Personalized Portrait Video Reconstruction Method Based on the Reptile Algorithm
Jin Zhang, Zhang We, Weimin Lei, Wanxin Zhang · 2025
To address the limitations of existing generative adversarial network-based reconstruction methods in terms of generalization ability, convergence speed, and personalized feature representation, this paper proposes a facial video reconstruction method that integrates a multi-task partitioning strategy, a meta-learning algorithm, and a continuous affine transformation motion modeling algorithm. Experiments conducted on the VoxCeleb2 and HDTF datasets demonstrate that the proposed method outperforms existing approaches such as MCNET in terms of SSIM, PSNR, LPIPS, and L1 metrics, while exhibiting strong generalization capabilities on unseen data.