Loading stylegan weights from pretrained!
Traceback (most recent call last):
File "/home/akira/DeltaEdit/scripts/inference_real.py", line 206, in
main(opts)
File "/home/akira/DeltaEdit/scripts/inference_real.py", line 189, in main
fake_delta_s = net(latent_s, delta_c)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/home/akira/DeltaEdit/./delta_mapper.py", line 60, in forward
x_coarse = torch.cat([s_coarse, torch.stack([c_coarse]*3, dim=1)], dim=2) #[b,3,1024]
RuntimeError: Sizes of tensors must match except in dimension 2. Expected size 2 but got size 1 for tensor number 1 in the list.
我自己上传了一张图片，为什么报这个错误了

Where is the code for the frame position of the training part above?

This is an amazing job！ I want to test inference with my own dataset, how should I do it?

I really like your work. I've encountered a problem using DeltaEdit for a different field, but during the training process, I encountered the following error:

Traceback (most recent call last): File "scripts/train.py", line 63, in main(opts) File "scripts/train.py", line 52, in main loss.backward() File "/root/autodl-tmp/conda/envs/Del/lib/python3.8/site-packages/torch/tensor.py", line 221, in backward torch.autograd.backward(self, gradient, retain_graph, create_graph) File "/root/autodl-tmp/conda/envs/Del/lib/python3.8/site-packages/torch/autograd/init.py", line 130, in backward Variable._execution_engine.run_backward( RuntimeError: Function FusedLeakyReLUFunctionBackward returned an invalid gradient at index 1 - got [4] but expected shape compatible with [512]

Is there any solution to this problem?

Do you have plans to share pre-trained models of LSUN cats, churches, and horses?

Thank you so much for open source. I have a few questions about some details:

1.I noticed that in your inference.py file, you use a file called fs3.npy. Could you explain what the data in this file means and how I can obtain it? I couldn't find any code in your repository for generating this file.

2.I also noticed that in your train.py file, you don't set a max_iters parameter, and only iterate through the dataset once. Does this mean that the delta mapper doesn't need multiple rounds of training?

3.Your open source code doesn't seem to include the code for cspace_ffhq_feat.npy and sspace_ffhq_feat.npy. Could you provide this code or tell me where I can find it?

4.When I tried editing the sample images you provided with the default parameters, I got the same results as shown in the paper for the provided texts. However, when I tried other texts, the results weren't very good. But when I used the text "Blue hair", I got a result that was close to the "Blue suit" in the paper. Have you encountered this situation before?

How to extend other editing requirements, such as generating a face with a single eyelid. I executed the following command :

python scripts/inference_real.py --target "face with big eyes","face with single eyelids"

but the output indicated that the third image, which was supposed to be a face with a single eyelid, had no effect.

I meet some problems when I try to edit the horse and cat, can you release training and inference code for the LSUN cat, church, horse please? Thank you very much

Congratulations on your paper being accepted by CVPR 2023！
Regarding Figure 2 in your paper,

1. can you provide the implementation code?
2. I tried hard to reproduce this result but failed. Can you elaborate on how the CLIP Delta feature is calculated here?

Hi, I am playing with your DeltaEdit project and when I try to diversify the edit hair color it does not work, this is the result of my run of editing the pink color.
origin：

result：

Hi, thanks for releasing this nice work!

If I want to inference with an image in the wild instead of FFHQ dataset, how can I extract the parameters in c/s/wspace_img_feat.npy file?
Could you share this preprocess script?

Thanks

Hello! When I run the CUDA_VISIBLE_DEVICES=0 python generate_codes.py --classname ffhq --samples 200000 I get the following error:

Traceback (most recent call last):
File "/.../DeltaEdit/generate_codes.py", line 118, in
netG = Generator(args.size, 512, 8).to(device)

RuntimeError: The NVIDIA driver on your system is too old (found version 11040). Please update your GPU driver by downloading and installing a new version from the URL: http://www.nvidia.com/Download/index.aspx Alternatively, go to: https://pytorch.org to install a PyTorch version that has been compiled with your version of the CUDA driver.

What should I do?

Thanks in advance.

Thank you very much for the open source, but I am having some problems with the training:
1, I looked at the training code, the whole training process is only traversing the training data set once (that is, 200000 + 58000)? But when I got to 201000, I finished the training directly without any error report, is this normal?

2, I noticed in your paper that you use the remaining 12,000 FFHQ images for evaluation, but I can't find the evaluation code and dataset, the parameter val_interval is not used, how should I evaluate my trained model?

          Hi, Why close my issue? Part of my problem is not yet solved, I hope to get your reply if possible.

Originally posted by @junyizeng in #3 (comment)

Hi, I tried many prompt to change the head pose, but all failed. Does it meet expectations?
If it meet expectations, how can I edit the head pose?

Hello, your work is excellent!
I have a question I'd like to ask you.
You know, we can get the new image after editing the updated style vector guided by three parameters of S (stylegan2 space vector), the original image obtained by e4e latent space vector latent (w+), and stylegan2 generator.
So can we get a new w+ vector based on these three parameters?

Switch to pip installation