We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CV

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Electrical Engineering and Systems Science > Image and Video Processing

Title: Towards Measuring Domain Shift in Histopathological Stain Translation in an Unsupervised Manner

Abstract: Domain shift in digital histopathology can occur when different stains or scanners are used, during stain translation, etc. A deep neural network trained on source data may not generalise well to data that has undergone some domain shift. An important step towards being robust to domain shift is the ability to detect and measure it. This article demonstrates that the PixelCNN and domain shift metric can be used to detect and quantify domain shift in digital histopathology, and they demonstrate a strong correlation with generalisation performance. These findings pave the way for a mechanism to infer the average performance of a model (trained on source data) on unseen and unlabelled target data.
Comments: 5 pages, 3 figures, 2 tables
Subjects: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
Journal reference: ISBI, 2022, pp. 1-5
DOI: 10.1109/ISBI52829.2022.9761411
Cite as: arXiv:2205.04368 [eess.IV]
  (or arXiv:2205.04368v1 [eess.IV] for this version)

Submission history

From: Zeeshan Nisar [view email]
[v1] Mon, 9 May 2022 15:18:12 GMT (10928kb,D)

Link back to: arXiv, form interface, contact.