ludles

Thank you for your reply! I still haven't got it working as expected. But it seems the key problem turns out to be how to "mimic" the resolution pyramid strategy...

I am working on [this dataset](https://doi.org/10.5281/zenodo.3874362) consisting of multimodal microscopy images. I have achieved reasonably good registration quality on it using an iterative approach through `SimpleElastix`. I'm just thinking it...

Sorry for the confusion! Here is an example. I used the left as the moving. 

> > > Hi! The new directory names 0, 1, ..., 999 cause trouble for using `torchvision.datasets.ImageFolder`, which sorts the names into 0, 1, 10, 100, ..., 999, different from...