`MILFeatures`: Calculate last-layer activations from MIL models

jamesdolezal / slideflow

Deep learning library for digital pathology, with both Tensorflow and PyTorch support.

https://slideflow.dev

GNU General Public License v3.0

234 stars 39 forks source link

`MILFeatures`: Calculate last-layer activations from MIL models #327

Closed jamesdolezal closed 8 months ago

jamesdolezal commented 8 months ago

Adds a new function sf.mil.generate_mil_features() that returns an MILFeatures object generated from extracting activation weights from the last layer of a TransMIL, AttentionMIL, CLAM_sb, or CLAM_mb model.

Written by @matte-esse and @CostanzaSiniscalchi