Fake data labels

[SanaNazari]

I have trained a new model with my dataset with Stylegan3 with labels. Where can I see the labels of the fake images? Or does it generate labels at all?

[hecoding]

gen_images.py will generate images, one class at a time. It can be controlled by the --class flag.

[49xxy]

我已经用我的带有标签的 Stylegan3 数据集训练了一个新模型。我在哪里可以看到假图像的标签？或者它是否会生成标签？

I want to know how your labels are set up, I want to get better generation quality with labels on my own dataset, but I don't know much about labels yet.

Sincerely hope to get your help.

[SanaNazari]

我已经用我的带有标签的 Stylegan3 数据集训练了一个新模型。我在哪里可以看到假图像的标签？或者它是否会生成标签？

I want to know how your labels are set up, I want to get better generation quality with labels on my own dataset, but I don't know much about labels yet.

Sincerely hope to get your help.

This is how the structure of my dataset.json looks like. I created a custom script to generate this. I divided my images into folders by class label . This structure can be later loaded by dataset_tool.py and it will generate another dataset.json file for the processed data.

Basic structure : {"labels":[["folder/image name", image label]]}

Example: {"labels": [["00000/ISIC_0024312.jpg", 0], ["00000/ISIC_0024324.jpg", 0], ["00000/ISIC_0024336.jpg", 0], ["00000/ISIC_0024337.jpg", 0]]}