MAISI mask autoencoder

[wachinger]

Hi,

Thanks for your great work with MAISI. The mask_generation_autoencoder has 8 input channels. What is the kind of data that is required as input? Would you have an example on how to use the mask autoencoder alone, without the diffusion part based on the provided data description for mask_pseudo_label*.nii.gz ?

"mask_generation_autoencoder_def": {
    "_target_": "monai.apps.generation.maisi.networks.autoencoderkl_maisi.AutoencoderKlMaisi",
    "spatial_dims": "@spatial_dims",
    "in_channels": 8,
    "out_channels": 125,

Thanks, Christian

[Can-Zhao]

@guopengf Hi Pengfei, could you help answer this question? Thank you!

[guopengf]

Hi @wachinger, you can use this function ldm_conditional_sample_one_mask to use the mask generation model alone. We will create a new tutorial on how to use a standalone mask generation diffusion model.

[mar-cry]

Hi @guopengf , i am also interested in the in_channels of the mask autoencoder, what is the data type of this input ?

[guopengf]

Please refer to the discussion here https://github.com/Project-MONAI/tutorials/discussions/2026#discussioncomment-14324998