GANInversion_with_ConsecutiveImgs

Official code for our ICCV paper: "From Continuity to Editability: Inverting GANs with Consecutive Images" https://arxiv.org/pdf/2107.13812.pdf

1. Build the environment with stylegan.yaml (Anaconda is required)
2. Compile FlowNet2 dependencies (correlation, resample, and channel norm layers).
Reference: https://github.com/phoenix104104/fast_blind_video_consistency.
3. Download the StyleGAN weight and FlowNet weight from: https://drive.google.com/file/d/1g2gp4tR0wAc6uG24qkt82pM3afD-vfBT/view?usp=sharing.
4. Python main.py