Skip to content

UNet2DS, UNet1D Models and Weights

Pre-release
Pre-release
Compare
Choose a tag to compare
@alexklibisz alexklibisz released this 11 Aug 13:01
· 17 commits to dev since this release

Keras models and weights for UNet2DS and UNet1D with some training artifacts. Model files were serialized in Python3.5 and might not work with other version of Python.

UNet2DS:

  • Weights from "Fast, Simple Calcium Imaging Segmentation with Fully Convolutional Networks"
  • Trained at this commit.
  • Validation F1 approximately 0.84. Neurofinder submission F1 = 0.569.
  • Including training metrics.

UNet1D:

  • Most recently-trained weights.
  • Trained at this commit.
  • Validation F2=0.81, recall=0.84, precision=0.71.
  • Including training metrics and a sample of predictions on validation data.