Framework

An overview of the proposed MCNet. It contains two designed modules to compensate the source facial feature map: (i) The implicit identity representation conditioned memory module (IICM) learns a global facial meta-memory bank, and an implicit identity representation from facial keypoint coordinates of the source image, which conditions on the query of the learned meta-memory bank, to obtain more structure-correlated facial memory to the warped source feature map for compensation; (ii) The memory compensation module (MCM) designs a dynamic cross-attention mechanism to perform a spatial compensation for the warped source feature map for the generation.

Samples

Abstract

Talking head video generation aims to animate a human face in a still image with dynamic poses and expressions using motion information derived from a target-driving video, while maintaining the person's identity in the source image. However, dramatic and complex motions in the driving video cause ambiguous generation, because the still source image cannot provide sufficient appearance information for occluded regions or delicate expression variations, which produces severe artifacts and significantly degrades the generation quality. To tackle this problem, we propose to learn a global facial representation space, and design a novel implicit identity representation conditioned memory compensation network, coined as MCNet, for high-fidelity talking head generation.~Specifically, we devise a network module to learn a unified spatial facial meta-memory bank from all training samples, which can provide rich facial structure and appearance priors to compensate warped source facial features for the generation. Furthermore, we propose an effective query mechanism based on implicit identity representations learned from the discrete keypoints of the source image. It can greatly facilitate the retrieval of more correlated information from the memory bank for the compensation. Extensive experiments demonstrate that MCNet can learn representative and complementary facial memory, and can clearly outperform previous state-of-the-art talking head generation methods on VoxCeleb1 and CelebV datasets.

Acknowledgements

Our code follows several awesome repositories such as DaGAN, FOMM, We appreciate them for making their codes available to public.
This research is supported in part by HKUST-SAIL joint research funding, the Early Career Scheme of the Research Grants Council (RGC) of the Hong Kong SAR under grant No. 26202321 and HKUST Startup Fund No. R9253.
The website template was borrowed from Jon Barron Mip-NeRF.