Question about Figure 2 in paper

[tanz63]

I'm wondering why in Figure 2, the input figure and reconstructed figure is different？since it's a schema for the conditional denoise training pipeline，isn't it should be the same figure ideally？

[rinongal]

It waws meant to convey that actual generated images may not be exactly the same as the input, but you're correct that in hindsight we should probably have used the same images with / without noise.