LivePortrait_ncnn

a naive example of LivePortrait infer by ncnn

Run mode :

1.Retargeting(lip open/eye open/head pose editing)
2.Source image driving by video/images

PS: infer time: 28s/driving image

Time profile

op type                    avg time(ms)      %
Conv3d(no simd/vulkan)     27168.23         96.24%
other op                   1059.54          3.76%
total time:                28227.77

Result

1.Retargeting

2.Source image driving by video

Reference

1.https://github.com/KwaiVGI/LivePortrait
2.https://github.com/Tencent/ncnn