Skip to content

v4.0.5

Choose a tag to compare

View all tags
@mrariden mrariden released this 25 Jun 19:59
· 61 commits to main since this release
v4.0.5
8ef8804
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Updates to Cellpose-SAM

  • resample added back to CLI for diameter changes. Note that the diameter parameter is only used for image rescaling and is not always necessary to provide in CPSAM compared to CP3.
  • default model weights are changed to torch.bfloat16 instead of previous 32-bit float data type. This decreases the model size by ~50% and speeds up runtime by ~40%, therefore helping with OOM issues. Changing back to the 32-bit model can be done during CellposeModel instantiation by setting use_bfloat16=False.
  • Improved 3d image dimension guessing
  • Fixed a bug that ran the network 4x more often than necessary when image size was <256 length or width.
  • Added some documentation
  • Updated notebooks
  • Some more comprehensive testing

Changelog: v4.0.4...v4.0.5

Assets 2
Loading