1. Overview
StyleGAN-FineTune-v1,v2,v3,v4 models for Oh-LoRA Project.
- These models are Fine-Tuned version of GenForce StyleGAN celeba_partial-256x256.
- Finally decided to use StyleGAN-FineTune-v3 (
stylegan_gen_fine_tuned_v3_ckpt_0005_gen.pth). - Detailed info (in Korean)
2. Model files & save path for usage
All models have the same save path for usage : 2025_04_08_OhLoRA/stylegan_and_segmentation/stylegan_modified/{model_file_name}.
- Generator / Discriminator
| Version | Model type | Model | Finally decided to use |
|---|---|---|---|
| StyleGAN-FineTune-v1 | Generator (GAN) | stylegan_gen_fine_tuned_v1.pth |
|
| StyleGAN-FineTune-v1 | Discriminator (GAN) | stylegan_dis_fine_tuned_v1.pth |
|
| StyleGAN-FineTune-v3 | Generator (C-VAE Decoder) | stylegan_gen_fine_tuned_v3_ckpt_0005_gen.pth |
✅ |
| StyleGAN-FineTune-v4 | Generator (GAN) | stylegan_gen_fine_tuned_v4.pth |
|
| StyleGAN-FineTune-v4 | Discriminator (GAN) | stylegan_dis_fine_tuned_v4.pth |
- Conv-NN for property score computation
| Version | Model type | Model | Finally decided to use |
|---|---|---|---|
| StyleGAN-FineTune-v2 | CNN (for property score computation) | stylegan_gen_fine_tuned_v2_cnn.pth |
✅ |
Inference Providers
NEW
This model isn't deployed by any Inference Provider.
🙋
Ask for provider support