Shalin Mehta

The question of chunk size came up again at the Donuts & Development, and that we should measure it. The chunk-size vs read/write benchmarks will guide both the users of...

I would change the abstract to `Pythonic and parallelizable I/O for N-dimensional imaging data with OME metadata". Other than that, LGTM. @talonchandler @ieivanov @edyoshikun please chime in too.

@ziw-liu We are trying two kinds of models: * Semantic segmentation (background vs uninfected vs infected) based on human annotation. * Regression of infection score based on engineered features. The...

Thanks @ziw-liu. **Re: regression models:** @Soorya19Pradeep The models that predict infection scores (phase IQR or skew of fluorescence) should be trainable with the current code. I suggest that you compute...

> That would be the "2.1D". I just checked and it has bit-rotted since mid-last year as we deprecated it for virtual staining. I can fix it if there's a...

@Soorya19Pradeep The annotations are constant along Z, right? It does make sense to use 6 slices of HSP90 and phase as input, and 1 slice of annotation as an output....

TODO: Implement tests via `hypothesis`

> The ultrack plugin does load the extra columns in its output CSVs as layer features so it can be used by the cluster plotter. That's interesting. Can this work?...

@ziw-liu @alishbaimran Given our offline discussion, here is the prioritization of features: 1) update data format and data module for efficiency + allow selection of channels to encode, 2) define...

@ziw-liu and @alishbaimran I think we can bypass the patchification step by chunking the Zarr store in `C*Y*X` sized chunks. The data chunked like this can be loaded fast enough...