Unsupervised Training of a Dynamic Context-Aware Deep Denoising Framework for Low-Dose Fluoroscopic Imaging

🚀 This repository contains the PyTorch implementation of a Two-step unsupervised dynamic context-aware denoising framework based on the recursive filter for Low-Dose fluoroscopic imaging.

• The code is an implementation of the algorithm proposed in the paper, and it will be uploaded once the paper is accepted.

Results

Low-dose	BM3D	NLM	N2N	N2V	N2C	Ours
WGAN-VGG	DnCNN	MCDnCNN	UDDN	EEDN	FastDVDNet	High-dose

🚀 Detailed Results and Demonstrations

This section provides detailed results and demonstrations of our model. It includes images from each algorithm as well as the corresponding video generated by combining these images. Through these results and demonstrations, a more comprehensive understanding of the performance and effectiveness of our model can be gained.

👉 Comparative denoising results (click to expand)

Updates

• First release: Coming soon!
• 😎 This repository is the official implementation of
  “Unsupervised Training of a Dynamic Context-Aware Deep Denoising Framework for Low-Dose Fluoroscopic Imaging.”
• The code and documentation are currently being organized and will be uploaded as soon as they are ready.
• Thank you for your patience and interest!

Preparation

Environment installation:

You can try Anaconda to setup the environment.

Anaconda

conda create -n UDCA-Net python=3.8.8
conda activate UDCA-Net
pip install torch==1.12.1+cu116 torchvision==0.13.1+cu116 torchaudio==0.12.1 --extra-index-url https://download.pytorch.org/whl/cu116
pip install -r requirements.txt

Training and Testing

Testing

python test.py --model [MODEL_NAME] 

Examples: python test.py --model UDCANet

Training

Training command:

python train.py --model UDCANet --patch_size [PATCH_SIZE] --n_frames [N_FRAMES] 

Examples: python train.py --model UDCANet --patch_size 120 --n_frames 5

Contact

if you have any questions, please file an issue or contact the author:

Sun-Young Jeon: sunyounge_@ewhain.net