Eduardo Hirata-Miyasaki comments

Results 37 comments of


                                            Eduardo Hirata-Miyasaki

Single-cell representation learning

@ziwen, I'm done testing the virtual staining end-to-end. I didn't run any of the representation learning. I really appreciate the new config files structure and CLI. This will work well...

Use roc auc instead of accuracy

@edyoshikun replace accuracy for ROC AUC for dynaclr

Always add indices to the sample

Is this meant to be merged? @ziw-liu or was it just for the hackathon?

update to iohub and zarr v3

We need Biahub and waveorder to upgrade to zarr v3 in tandem https://github.com/czbiohub-sf/biahub/pull/104

vae with temporal matching with positive pair

[vae_tmapping.pdf](https://github.com/user-attachments/files/23556407/vae_tmapping.pdf)

vae with temporal matching with positive pair

EH: take a look at the graph and check that we are doing the backprop throught the same path.

Issues with VSNeuromast Inference

Here, I am referring to the examples [VSNeuromast](https://github.com/mehta-lab/VisCy/blob/main/examples/virtual_staining/VS_model_inference/demo_vsneuromast.py). We should clarify that you need to run the preprocessing explicitly. You can't run the code with any other normalization other than...

FID code from Dihan

Replace the FID with torchmetrics https://lightning.ai/docs/torchmetrics/stable/image/frechet_inception_distance.html

Organize segmentations, tracking, annotations, and embeddings with DynaCell zarr stores

The suggestion above resembles the Spatialdata format. Let's evaluate if this is flexible and performant for our model training. - https://spatialdata.scverse.org/en/stable/tutorials/notebooks/notebooks/examples/intro.html - https://liveimagetrackingtools.org/geff/latest/specification/#extra-attributes