cedricoeldorf/conditionalstylegan

Conditional implementation for NVIDIA's StyleGAN architecture

Instruction to myself

css

Jekyll rebuilds anything in the build folder, thus overwriting any changes. For custom CSS, access assets -> css -> main.scss

conditionalstylegan's Issues

Is it possible to interpolate the images of different classes?

Hi!

I am wondering if I could make latent walk traversing different classes, for example, the vector with label 1 and the vector with label 2.

Thanks!

Do you share the trained model somewhere?

Or it must be retrained from scratch?

where is the file "mypickle.pickle" referenced in the code?

a

mypickle.pickle no found

in dataset_tool.py , all_data = unpickle('../data/mypickle.pickle') line = 300

Please share the code for "Experiment 3"

Looks like the repository has the code for "Experiment 2: Object Classification Based
Conditions", but not for "Experiment 3: ResNet Feature Conditions".

re-create experiment, can not find labels (conditions)

Hi, I would like to re-create your experiment, but I can not find your labels (conditions) for your dataset. Are they publicly available?

Custom Input

How to customize input, in a sense that the input is not a onehot vector?

Thanks

Error : module 'tensorflow' has no attribute 'Dimension'

When running the pretrained_example.py file, I have

Traceback (most recent call last):
File "C:/Users/CAO/Desktop/ConditionalStyleGAN/pretrained_example.py", line 15, in
import dnnlib.tflib as tflib
File "C:\Users\CAO\Desktop\ConditionalStyleGAN\dnnlib\tflib_init_.py", line 8, in
from . import autosummary
File "C:\Users\CAO\Desktop\ConditionalStyleGAN\dnnlib\tflib\autosummary.py", line 31, in
from . import tfutil
File "C:\Users\CAO\Desktop\ConditionalStyleGAN\dnnlib\tflib\tfutil.py", line 34, in
def shape_to_list(shape: Iterable[tf.Dimension]) -> List[Union[int, None]]:
AttributeError: module 'tensorflow' has no attribute 'Dimension'

Stylemixing

Hi,
How can I perform Style-mixing with 2/more uploaded images?
Greetings!

How do you remove text-based images?

Amazing work！But how do you remove the text-based images from LLD dataset? Just remove them manually?

How to exactly configure the pickle file? Error in dataset_tool.py

I have tried to configure the pickle file a few ways now, though I keep getting this error in dataset_tool.py:

File "dataset_tool.py", line 372, in create_from_images
    onehot = np.zeros((conditions.size, np.max(conditions) + 1), dtype=np.float32)
TypeError: must be str, not int

I have been referencing this line for formatting the pickle dictionary:

mypickle = {"Filenames": list_of_file_paths, "Labels": class_condition_labels}

I have a list of file names that are being correctly accessed it seems, then the "Labels" field is filled with a nested array of class condition labels like this [ ['A'], ['C'], ['E'], ... ] which I figure each nested array represents the class labels assigned to each file at the same index in the file name list. This yields this error:

File "dataset_tool.py", line 372, in create_from_images
    onehot = np.zeros((conditions.size, np.max(conditions) + 1), dtype=np.float32)
TypeError: can only concatenate list (not "int") to list

Thus, I am having a hard time diagnosing this error. Am I not formatting something correctly? Thanks!

Stylemixing

Generate New Images after training

Hi!
I just finished the training.Now I wanted to Generate new Images from saved weights by giving :

1. Latent Vector
2. Condition

I am able to generate new images without giving any condition but I wanted to generate new Images followed by Condtion .How do I do ?

Thanks in Advance!
Ash