Hi @dylran ,
I noticed that performance gaps between Diffuse-Denoise-Count, which has been proposed in the first version of this paper on arXiv https://arxiv.org/pdf/2303.12790v1, and CrowdDiff, which is in the third version of this paper https://arxiv.org/pdf/2303.12790v3, are significant.
Just wondering, what are the differences between these two models?
Looking forward to your reply.
Many thanks,
Yiming