最近有流行了深度学习算法的实现

本文采取最新的算法，实现了网络的设计，并且展示了数据结果

CVPR2022 | 曾经火爆全网的算法！升级版来袭，支持卡通形象！

作者开源了现有的代码

本文主要参考下面的网址做的实验GitHub - harlanhong/CVPR2022-DaGAN: Official code for CVPR2022 paper: Depth-Aware Generative Adversarial Network for Talking Head Video Generation

首先介绍了一种自监督几何学习方法，可以自动从人脸视频中恢复密集的 3D 几何（即深度），而不需要任何昂贵的 3D 标注数据。基于学习到的密集深度图，我们进一步建议利用它们来估计捕捉人头关键运动的稀疏面部关键点。以更密集的方式，深度还用于学习 3D 感知的跨模态（即外观和深度）注意力，以指导生成运动场以扭曲源图像表示。所有这些贡献构成了一个新颖的深度感知生成对抗网络（DaGAN），用于talking head 生成。进行的大量实验表明，我们提出的方法可以生成高度逼真的人脸，并在看不见的人脸上取得显着效果

图像下载于百度网络

如果有侵权 请联系坐着删除

谢谢

主要的方法

我认为就是让驱动图像和原始图像得到深度图 然后对深度图抽取关键点，那么关键点可以进行映射（那么两幅图原始图的映射关系也就有了）

根据原始图像的森赌徒 可以得到其密度图（或者称之为注意力机制图） 进而通过密度图经过映射，可以得到解码后出来的图形，加码后出来的图像的姿态就和驱动图像的姿态一样了 

代码

python demo.py  --config ./config/vox-adv-256.yaml  --checkpoint DaGAN_vox_adv_256.pth.tar --relative --adapt_scale --kp_num 15 --generator DepthAwareGenerator

Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
Lossy conversion from float32 to uint8. Range [0, 1]. Convert image to uint8 prior to saving to suppress this warning.
MoviePy - Building file res5528.mp4.gif with imageio.
t:  64%|███████████████████████████████████████████▏                       | 107/166 [00