VisCy 0.1.0

This is first release of VisCy, the machine learning pipeline to train and deploy computer vision models for single-cell phenotyping.

With 0.1.0 the following key features are available:

• Training, evaluation, inference, and deployment of virtual staining models based on 2D Residual U-Net, 2.5D U-Net, 3D U-Net, and UNeXt2 architectures
• Data module implementations for HCS OME-Zarr datasets, as well public test datasets like LiveCell and CTMC v1.
• Composing datasets and transformations for training and validation
• Distributed (DDP) training

The weights of the virtual staining models reported in the preprint can be found in the binaries section below.

What's Changed

• Migration from microDL by @ziw-liu in #1
• Update README by @ziw-liu in #22
• Bump iohub version by @ziw-liu in #25
• readme + dependencies tested with python 3.10 by @mattersoflight in #30
• Demo notebooks by @ziw-liu in #29
• 3D augmentation and 2.1D U-Net by @ziw-liu in #27
• Fix datamodule by @ziw-liu in #28
• Improve augmentation by @ziw-liu in #31
• Fix inference by @ziw-liu in #32
• viscy -> VisCy by @mattersoflight in #34
• Update readme.md typo for cd VisCy by @edyoshikun in #41
• 2.1D upscale decoder by @ziw-liu in #37
• dlmbl 2023 archive by @mattersoflight in #44
• Fix center slice metrics for 3D output by @ziw-liu in #51
• Configure the number of image samples logged at each epoch and batch by @ziw-liu in #49
• Example workflow by @ziw-liu in #45
• Project icon by @ziw-liu in #38
• Fix predicting new channels in an existing store by @ziw-liu in #57
• Document data methods by @ziw-liu in #50
• Visualize feature maps by @ziw-liu in #53
• Baseline 3D-LUNeXt by @ziw-liu in #58
• Preprocess CLI and source scaling during prediction by @ziw-liu in #59
• Configurable augmentations by @ziw-liu in #61
• Bump dependencies and update documentation by @ziw-liu in #64
• checkpoint as a model config parameter for warmup cosine learning rates by @edyoshikun in #66
• Masked autoencoder pre-training for virtual staining models by @ziw-liu in #67
• Filter empty detections in labels by @ziw-liu in #74
• Add CITATION.cff by @ziw-liu in #79
• Fix 3D to 2D prediction with UNeXt2 model by @ziw-liu in #80
• 2D FCMAE by @ziw-liu in #71
• Rename UNeXt2 by @ziw-liu in #84
• Test on Python 3.12 by @ziw-liu in #88
• Add preprint reference to README by @ziw-liu in #85
• Fix architecture name in network diagram script by @ziw-liu in #86
• Add the scale metadata to the output_stores by @edyoshikun in #89
• bumping to cellpose 3 by @edyoshikun in #92
• Scale metadata handling for positions by @edyoshikun in #93
• Demo for VSCyto2D and VSCyto3D by @edyoshikun in #94
• Fix demos on other platforms by @ziw-liu in #95

Full Changelog: https://github.com/mehta-lab/VisCy/commits/v0.1.0