Skip to content

This repository is a PyTorch implementation for PET tumor generation from benign PET and tumor mask

Notifications You must be signed in to change notification settings

Saint-lsy/Mask2PET

This branch is 1 commit ahead of o0t1ng0o/Mask2PET:main.

Folders and files

NameName
Last commit message
Last commit date

Latest commit

72e8b96 · Dec 15, 2024

History

3 Commits
Dec 15, 2024
Jul 28, 2024
Dec 15, 2024
Dec 15, 2024
Jul 28, 2024
Jul 28, 2024
Dec 15, 2024
Jul 28, 2024
Dec 15, 2024
Jul 28, 2024
Dec 15, 2024
Jul 28, 2024

Repository files navigation

Mask2PET

This repository is a PyTorch implementation for PET tumor generation from benign PET and tumor mask. We build this code upon medfusion. You can completely follow the instruction in medfusion.

Dataset

We use the HECKTOR 2021 dataset, please download from this link (https://www.aicrowd.com/challenges/miccai-2021-hecktor). Please check the path of data in ./launch/train.sh and ./launch/test.sh

Data preprocessing

In this work, we use the VQ-GAN as an encoder to encode the PET. If you want to customize this model on your dataset, please train VQ-GAN on your dataset first. https://github.com/CompVis/taming-transformers.

We provide the pre-trained VQ-GAN model at https://huggingface.co/Valentina007/Mask2PET, please download them and place them to ./pretrained_models .

Pre-trained model of our model

You can download the well-trained model at https://huggingface.co/Valentina007/Mask2PET.

Training

sh ./launch/train.sh

inference

sh ./launch/test.sh

About

This repository is a PyTorch implementation for PET tumor generation from benign PET and tumor mask

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 94.3%
  • Python 5.7%