You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Is your feature request related to a problem? Please describe.
Image generation with MAISI
Describe the solution you'd like
Afaiu, currently, MAISI can "only" generate volumes conditionally on controlnet_cond. These are relative size changes of certain anatomical labels. But it cannot generate a volume from a concrete labelmap that is given.
Describe alternatives you've considered
In ControlNet's forward() method, I did not see an alternative that would allow embedding a concrete labelmap. The Diffusion Model requires embeddings from ControlNet, and cannot accept concrete labelmaps either.
Additional context Figure 2 in the MAISI paper also seems to indicate that ControlNet can only accept "task-specific conditions", and "Diffusion Model" depends on ControlNet embeddings.
The text was updated successfully, but these errors were encountered:
Is your feature request related to a problem? Please describe.
Image generation with MAISI
Describe the solution you'd like
Afaiu, currently, MAISI can "only" generate volumes conditionally on
controlnet_cond
. These are relative size changes of certain anatomical labels. But it cannot generate a volume from a concrete labelmap that is given.Describe alternatives you've considered
In ControlNet's
forward()
method, I did not see an alternative that would allow embedding a concrete labelmap. The Diffusion Model requires embeddings from ControlNet, and cannot accept concrete labelmaps either.Additional context
Figure 2 in the MAISI paper also seems to indicate that ControlNet can only accept "task-specific conditions", and "Diffusion Model" depends on ControlNet embeddings.
The text was updated successfully, but these errors were encountered: