Problem consultation

[Gbone3176]

hello~ Our dataset does not have a corresponding semantic segmentation picture, so can we still use your network and weights for training? thanks!

[pmh9960]

Thank you for your interest.

Unfortunately, our approach only supports image-label paired datasets. I’m not sure if this applies to your dataset, but it may not work well. Nevertheless, the advanced training techniques of joint diffusion models with unpaired datasets could be a promising research direction.

Best,
Minho