Thanks to visit codestin.com
Credit goes to github.com

Skip to content

[ICCV 2025] This is the official PyTorch codes for the paper: "DiT4SR: Taming Diffusion Transformer for Real-World Image Super-Resolution"

Notifications You must be signed in to change notification settings

jinlovespho/DiT4SR

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

52 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

⚒️ Training Preparation

1. Environment

conda create -n dit4sr python=3.9 -y
conda activate dit4sr

2. Installation

Install the libraries in the order listed below.

pip install torch==2.2.2 torchvision==0.17.2 torchaudio==2.2.2 --index-url https://download.pytorch.org/whl/cu121
pip install -r requirements.txt
pip install numpy==1.26.3 --no-deps
pip install pyiqa==0.1.14.1 --no-deps 
cd detectron2 
pip install -e . --no-deps --no-build-isolation --config-settings editable_mode=compat
cd ../testr 
pip install -e . --no-deps --no-build-isolation --config-settings editable_mode=compat
pip install cloudpickle --no-deps

3. Download Weights and Dataset

3-1. Training Dataset

3-2. Pretrained Weights

  • SD3 weights
# First login enter your HF access token to download SD3 weights
huggingface-cli login   

# Then run the bash file
bash download_bash/download_sd3.sh 
  • DiT4SR weights
bash download_bash/download_dit4sr.sh 
  • TESTR weights
bash download_bash/download_testr.sh 

🔥 Training Recipe

Before training, modify the training script and training config.

  • training script -> for setting cuda and GPU
  • training config -> for setting data_path, training_batch_size, etc

Stage1 Training

# stage1 training 
bash run_scripts/train/JIHYE_train_stage1_dit4sr.sh

Stage2 Training

# stage2 training 
bash run_scripts/train/JIHYE_train_stage2_testr.sh

Stage3 Training

# stage3 training 
bash run_scripts/train/JIHYE_train_stage3_dit4sr_testr.sh

About

[ICCV 2025] This is the official PyTorch codes for the paper: "DiT4SR: Taming Diffusion Transformer for Real-World Image Super-Resolution"

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 93.6%
  • Cuda 3.6%
  • C++ 1.8%
  • Shell 0.5%
  • JavaScript 0.2%
  • HTML 0.2%
  • Other 0.1%