Some GAN inversion methods I wrote recently. Mention that 1-4 are all wrote based on pi-GAN.
1\
multi_view_inversion.py is an optimization based inversion on W space which not only used front image but also left_30 & right_30 images as a 3D information supplenment.The data is selected and processed from MEAD.
2\
inversion_on_w_space.py is a normal optimization based inversion on W space.
3
inversion_on_z.py is a simple but efficent script to inverse an image quickly on z latent space.
4
train_single_image_encoder.py is part of the normal learning based inversion pipeline which used to train a single image encoder. This encoder has the same architecture with discriminator in pi-GAN which can encode an single image or a batch of images into latent code.