search

zhanglonghao1992 / One-Shot_Free-View_Neural_Talking_Head_Synthesis

Pytorch implementation of paper "One-Shot Free-View Neural Talking-Head Synthesis for Video Conferencing"
Other
782 stars 143 forks source link

crop source #73

Open yja1 opened 1 year ago

yja1 commented 1 year ago

请问对source img有额外的预处理么？或者是训练过程加了额外的crop之类的数据增广？ （因为测试过程发现source img采用不同方式crop,drive的结果差不多。这是怎么做到的呢？）