tangjiapeng / DiffuScene

[CVPR 2024] DiffuScene: Denoising Diffusion Models for Generative Indoor Scene Synthesis
Other
225 stars 21 forks source link

why did the checkpoint is trained for each room type? #36

Open gyula-coder opened 4 months ago

gyula-coder commented 4 months ago

I‘d like to know the reason why why did you train a checkpoint for each room type? Is this the best approach after some trial and error?

Another way, room type is added as a condition to Unet and training a checkpoint for all room type. Is it possible?

Thanks for any help

tangjiapeng commented 3 months ago

To ensure fair comparison with previous methods, we train different diffusion models for different room types.

But yeah, it's possible to train a unified diffusion model with room type condition