Skip to content

Latest commit

 

History

History
229 lines (194 loc) · 10 KB

README.md

File metadata and controls

229 lines (194 loc) · 10 KB

FontDiffuser: One-Shot Font Generation via Denoising Diffusion with Multi-Scale Content Aggregation and Style Contrastive Learning

FontDiffuser_LOGO

arXiv preprint Gradio demo Homepage Code

🔥 Model Zoo 🛠️ Installation 🏋️ Training📺 Sampling📱 Run WebUI

🌟 Highlights

Vis_1 Vis_2

  • We propose FontDiffuser, which can generate unseen characters and styles and can be extended to cross-lingual generation, such as Chinese to Korean.
  • FontDiffuser excels in generating complex characters and handling large style variations. And it achieves state-of-the-art performance.
  • The generated results by FontDiffuser can be perfectly used for InstructPix2Pix for decoration, as shown in thr above figure.
  • We release the 💻Hugging Face Demo online! Welcome to Try it Out!

📅 News

  • 2024.01.27: The training of phase 2 is released.
  • 2023.12.20: Our repository is public! 👏🤗
  • 2023.12.19: 🔥🎉 The 💻Hugging Face Demo is public! Welcome to try it out!
  • 2023.12.16: The gradio app demo is released.
  • 2023.12.10: Release source code with phase 1 training and sampling.
  • 2023.12.09: 🎉🎉 Our paper is accepted by AAAI2024.
  • Previously: Our Recommendations-of-Diffusion-for-Text-Image repo is public, which contains a paper collection of recent diffusion models for text-image generation tasks. Welcome to check it out!

🔥 Model Zoo

Model chekcpoint status
FontDiffuer GoogleDrive / BaiduYun:gexg Released
SCR GoogleDrive / BaiduYun:gexg Released

🚧 TODO List

  • Add phase 1 training and sampling script.
  • Add WebUI demo.
  • Push demo to Hugging Face.
  • Add phase 2 training script and checkpoint.
  • Add the pre-training of SCR module.
  • Combined with InstructPix2Pix.

🛠️ Installation

Prerequisites (Recommended)

  • Linux
  • Python 3.9
  • Pytorch 1.13.1
  • CUDA 11.7

Environment Setup

Clone this repo:

git clone https://github.com/yeungchenwa/FontDiffuser.git

Step 0: Download and install Miniconda from the official website.

Step 1: Create a conda environment and activate it.

conda create -n fontdiffuser python=3.9 -y
conda activate fontdiffuser

Step 2: Install related version Pytorch following here.

# Suggested
pip install torch==1.13.1+cu117 torchvision==0.14.1+cu117 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu117

Step 3: Install the required packages.

pip install -r requirements.txt

🏋️ Training

Data Construction

The training data files tree should be (The data examples are shown in directory data_examples/train/):

├──data_examples
│   └── train
│       ├── ContentImage
│       │   ├── char0.png
│       │   ├── char1.png
│       │   ├── char2.png
│       │   └── ...
│       └── TargetImage.png
│           ├── style0
│           │     ├──style0+char0.png
│           │     ├──style0+char1.png
│           │     └── ...
│           ├── style1
│           │     ├──style1+char0.png
│           │     ├──style1+char1.png
│           │     └── ...
│           ├── style2
│           │     ├──style2+char0.png
│           │     ├──style2+char1.png
│           │     └── ...
│           └── ...

Training Configuration

Before running the training script (including the following three modes), you should set the training configuration, such as distributed training, through:

accelerate config

Training - Pretraining of SCR

Coming Soon ...

Training - Phase 1

sh train_phase_1.sh
  • data_root: The data root, as ./data_examples
  • output_dir: The training output logs and checkpoints saving directory.
  • resolution: The resolution of the UNet in our diffusion model.
  • style_image_size: The resolution of the style image, can be different with resolution.
  • content_image_size: The resolution of the content image, should be the same as the resolution.
  • channel_attn: Whether to use the channel attention in the MCA block.
  • train_batch_size: The batch size in the training.
  • max_train_steps: The maximum of the training steps.
  • learning_rate: The learning rate when training.
  • ckpt_interval: The checkpoint saving interval when training.
  • drop_prob: The classifier-free guidance training probability.

Training - Phase 2

After the phase 2 training, you should put the trained checkpoint files (unet.pth, content_encoder.pth, and style_encoder.pth) to the directory phase_1_ckpt. During phase 2, these parameters will be resumed.

sh train_phase_2.sh
  • phase_2: Tag to phase 2 training.
  • phase_1_ckpt_dir: The model checkpoints saving directory after phase 1 training.
  • scr_ckpt_path: The ckpt path of pre-trained SCR module. You can download it from above 🔥Model Zoo.
  • sc_coefficient: The coefficient of style contrastive loss for supervision.
  • num_neg: The number of negative samples, default to be 16.

📺 Sampling

Step 1 => Prepare the checkpoint

Option (1) Download the checkpoint following GoogleDrive / BaiduYun:gexg, then put the ckpt to the root directory, including the files unet.pth, content_encoder.pth, and style_encoder.pth.
Option (2) Put your re-training checkpoint folder ckpt to the root directory, including the files unet.pth, content_encoder.pth, and style_encoder.pth.

Step 2 => Run the script

(1) Sampling image from content image and reference image.

sh script/sample_content_image.sh
  • ckpt_dir: The model checkpoints saving directory.
  • content_image_path: The content/source image path.
  • style_image_path: The style/reference image path.
  • save_image: set True if saving as images.
  • save_image_dir: The image saving directory, the saving files including an out_single.png and an out_with_cs.png.
  • device: The sampling device, recommended GPU acceleration.
  • guidance_scale: The classifier-free sampling guidance scale.
  • num_inference_steps: The inference step by DPM-Solver++.

(2) Sampling image from content character.
Note Maybe you need a ttf file that contains numerous Chinese characters, you can download it from BaiduYun:wrth.

sh script/sample_content_character.sh
  • character_input: If set True, use character string as content/source input.
  • content_character: The content/source content character string.
  • The other parameters are the same as the above option (1).

📱 Run WebUI

(1) Sampling by FontDiffuser

gradio gradio_app.py

Example:

(2) Sampling by FontDiffuser and Rendering by InstructPix2Pix

Coming Soon ...

🌄 Gallery

Characters of hard level of complexity

vis_hard

Characters of medium level of complexity

vis_medium

Characters of easy level of complexity

vis_easy

Cross-Lingual Generation (Chinese to Korean)

vis_korean

💙 Acknowledgement

Copyright

Citation

@inproceedings{yang2024fontdiffuser,
  title={FontDiffuser: One-Shot Font Generation via Denoising Diffusion with Multi-Scale Content Aggregation and Style Contrastive Learning},
  author={Yang, Zhenhua and Peng, Dezhi and Kong, Yuxin and Zhang, Yuyi and Yao, Cong and Jin, Lianwen},
  booktitle={Proceedings of the AAAI conference on artificial intelligence},
  year={2024}
}

⭐ Star Rising

Star Rising