GenerativeModels

GenerativeModels copied to clipboard

Implement Score-MRI

3

Score-MRI ([paper](https://arxiv.org/abs/2110.05243), [code](https://github.com/HJ-harry/score-MRI)) is one of the simplest MRI reconstruction algorithms that can be implemented with a general pre-trained checkpoint. Once a diffusion model checkpoint is ready in the model...

hyungjin-chung

training time

1

Hello, thank you for sharing this wonderful package. Hello, I would like to ask, how many GPU cards did you use and how long did it take to train the...

hanfeng0409

Now that concat-based conditioning is supported in the infereres we should update tutorials to use the new infererers. AFAIK it is these two tutorials: https://github.com/Project-MONAI/GenerativeModels/blob/main/tutorials/generative/2d_super_resolution/2d_stable_diffusion_v2_super_resolution.ipynb https://github.com/Project-MONAI/GenerativeModels/blob/main/tutorials/generative/image_to_image_translation/tutorial_segmentation_with_ddpm.ipynb

marksgraham

Potential feature: One-shot generative models based on single training image

2

For use-cases where there are very few training images (e.g. perhaps rare diseases), it could be nice to augment the data using one-shot generative models. Could this be something of...

RichardObi

One of the big disadvantages of the FID is that to obtain meaningful results, the total number of samples should be equal/bigger than the latent space of the model used...

JessyD

Implement #293

Warvito

Add tutorial on generative models trained using federated scenario

1

Create a tutorial that uses monai.fl tools to train a generative models. It can use any FL tools for it (reccomend start with NVFlare), but it should use MONAIAlgo or...

Warvito

Similar to [RoentGen: Vision-Language Foundation Model for Chest X-ray Generation](https://arxiv.org/pdf/2211.12737.pdf), it would be good to be able to easily download the Stable diffusion model and then finetune it into a...

Warvito

Implements #154

Warvito

In order to make simpler the use of non-imaging data as conditioning, some new transformations could be added. For example, for text we have been using custom transformations to load...

Warvito