BillyXYB
commented
2 years ago Please check the inversion network and the generation model should be correctly paired. Did you use the inversion for celebA but generator for ffhq?
Open ms-ufad opened 2 years ago
BillyXYB
commented
2 years ago Please check the inversion network and the generation model should be correctly paired. Did you use the inversion for celebA but generator for ffhq?
为什么编辑一张celeba的图片，生成的图跟乱码似的。完全看不起人脸，是什么参数我没设置对吗?运行步骤如下： python3 dual_space_encoder_test.py --dataset_dir image/ --output_dir output --checkpoint_path best_model.pt
python our_interfaceGAN/edit_all_inversion_ffhq.py --ckpt ./out/transeditor_ffhq/checkpoint/790000.pt --num_region 1 --num_trans 8 --attribute_name gender --z_latent output/encoder_inversion/ffhq_encode/encoded_z.npy --p_latent output/encoder_inversion/ffhq_encode/encoded_p.npy --num_sample 1000