Disentangled representations of microscopy images

Jacopo Dapueto, Vito Paolo Pastore, Nicoletta Noceti, Francesca Odone · 2025

Read Disentangled representations of microscopy images by Jacopo Dapueto, Vito Paolo Pastore, Nicoletta Noceti, Francesca Odone (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-20649.txt.

Description

Jacopo Dapueto and co-authors teach microscope image models to show their reasons in a 2025 paper. Deep networks sort microscope images well, yet their choices stay opaque, a real bar in medicine and science. The team uses disentangled learning, which forces the model to store each visual factor apart.

Their twist trains the clean representation on synthetic images, then moves it to real data. Tests ran on three unlike sets: plankton, yeast cell pockets, and human cells. Across all three, the method kept accuracy high while giving factors a human can read. The authors frame that trade as the right one for lab use.

Catalog record

CREATORJacopo Dapueto, Vito Paolo Pastore, Nicoletta Noceti, Francesca Odone
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE34 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-20649

Browse the full library · About the Tower of Babel