To generate latent for new data, what shoud be revise under ../../GAN2Shape/gan2shape/stylegan2/stylegan2-pytorch/project_dist.sh

[cynthia-you]

Dr.Pan ,thanks for sharing your work.

I'm trying to reconstruct the 3D shape or info of old bulidng according to your repo. The inference of church(you offered) works well . But for new data , i'm a little bit confused, could you please give me some hint.

In the bash of ../../GAN2Shape/gan2shape/stylegan2/stylegan2-pytorch/project_dist.sh , you write: python projector.py \ --ckpt checkpoints/stylegan2-synface40k-config-e.pt --size 128 \ --step 1001 --channel_multiplier 1 --synface \ --with_vgg --lr 0.1 --noise 0.05 --no_noise \ --list_path /home/xgpan/projects/GAN/unsup3d/data/synface/train/img_list_200.txt \ --img_root /home/xgpan/projects/GAN/unsup3d/data/synface/train/image \ --save_path project/synface_train_40k \ --batch_size 25 --distributed

Frist , I found that the checkpoints/stylegan2-synface40k-config-e.pt donest exist, so i revised the sh like: !python /content/drive/MyDrive/GAN2Shape/gan2shape/stylegan2/stylegan2-pytorch/projector.py \ --ckpt /content/drive/MyDrive/GAN2Shape/checkpoints/stylegan2/stylegan2-synface-config-e.pt --size 128 \ --step 1001 --channel_multiplier 1 --synface \ --with_vgg --lr 0.1 --noise 0.05 --no_noise \ --list_path /content/drive/MyDrive/GAN2Shape/my_data/train/img_list.txt \ --img_root /content/drive/MyDrive/GAN2Shape/my_data/train/depth \ --save_path /content/drive/MyDrive/GAN2Shape/my_data/latent \ --batch_size 1

Actually, i have no idea of the exact checkpoints i should revise. As for the input size, i revised the from 128 to 64 to 256.

Since I don't know which ckpts to use, I've been arranging and combining the ckpts(/checkpoints/stylegan2/, wich your scripts/dounlad.sh mentioned) and input sizes... Could you please tell how to revise? thanx~~~~

[XingangPan]

For church, you should use checkpoint at checkpoints/stylegan2/stylegan2-church-config-f.pt (--size 256 --channel_multiplier 2). You can find these settings at https://github.com/XingangPan/GAN2Shape/blob/main/configs/church.yml

[cynthia-you]

For church, you should use checkpoint at checkpoints/stylegan2/stylegan2-church-config-f.pt (--size 256 --channel_multiplier 2). You can find these settings at https://github.com/XingangPan/GAN2Shape/blob/main/configs/church.yml

Thanks for you reply. But it seems that the vgg.pt is not compatible to the church.yml !python /content/drive/MyDrive/GAN2Shape/gan2shape/stylegan2/stylegan2-pytorch/projector.py \ --ckpt /content/drive/MyDrive/GAN2Shape/checkpoints/stylegan2/stylegan2-church-config-f.pt --size 256 \ --step 1001 --channel_multiplier 2 --synface \ --with_vgg --lr 0.1 --noise 0.05 --no_noise \ --list_path /content/drive/MyDrive/GAN2Shape/my_data/train/img_list.txt \ --img_root /content/drive/MyDrive/GAN2Shape/my_data/train/depth \ --save_path /content/drive/MyDrive/GAN2Shape/my_data/latent \ --batch_size 1

the error: Setting up Perceptual loss... Loading model from: /content/drive/MyDrive/GAN2Shape/gan2shape/stylegan2/stylegan2-pytorch/lpips/weights/v0.1/vgg.pth Traceback (most recent call last): File "/content/drive/MyDrive/GAN2Shape/gan2shape/stylegan2/stylegan2-pytorch/projector.py", line 189, in <module> model='net-lin', net='vgg', use_gpu=True, gpu_ids=[rank] File "/content/drive/MyDrive/GAN2Shape/gan2shape/stylegan2/stylegan2-pytorch/lpips/__init__.py", line 21, in __init__ self.model.initialize(model=model, net=net, use_gpu=use_gpu, colorspace=colorspace, spatial=self.spatial, gpu_ids=gpu_ids) File "/content/drive/MyDrive/GAN2Shape/gan2shape/stylegan2/stylegan2-pytorch/lpips/dist_model.py", line 75, in initialize self.net.load_state_dict(torch.load(model_path, **kw), strict=False) File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 386, in load return _load(f, map_location, pickle_module, **pickle_load_args) File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 573, in _load result = unpickler.load() File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 536, in persistent_load deserialized_objects[root_key] = restore_location(obj, location) File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 409, in restore_location result = map_location(storage, location) TypeError: 'int' object is not callable

I checked the projector.py the argument must be: parser.add_argument('--synface', action='store_true')

Could u tell me what can i do to fix it

[jojoWd]

no /GAN2Shape/gan2shape/stylegan2/stylegan2-pytorch/project_dist.sh
no projector.py Please let me get where to load those files?