ELEPHANT server v0.2.0

Updates

Add XYZ flip operations in data augmentation
Change MIN_AREA_ELLIPSOID from 20 to 9
Automatically initialize a new model with the pretrained parameters
Make detection algorithm compatible with 2D
Make flow algorithm compatible with 2D
Make export function compatible with 2D
Update loss functions for detection (See details in the paper)
Add tests for GitHub Actions
Add get_gpus() for getting GPU info from the client software
Implement upload endpoint for the client software
Implement dataset endpoint for the client software
Add environment.yml to reproduce the conda environment
Implement logger that can be shared with the client software
Update Colab notebook
Use averaged training losses in log
log_interval can be set as a parameter
Reorganize Redis state management
Reorganize cancel behavior
Stop using probability to refine spot radii
Move keep_axials from model to dataset
Upgrade tensorflow to 2.4.0
Organize model reset strategy
- Versatile -> latest versatile model
- Default -> self-supervised (detection) or random (flow)
- URL -> load from the specified URL
Update README