Zpět na vyhledávání

Bc. Martin Krebs

Bachelor's thesis

Interpretation techniques for deep neural networks in digital histopathology

Interpretation techniques for deep neural networks in digital histopathology
Abstract:
Skupina RationAI natrénovala konvolučnú neurónovú sieť, ktorá predpovedá výskyt rakoviny prostaty v digitálnych snímkach tkaniva. Naším cieľom je vedieť dostatočne rýchlo vysvetliť správanie tejto siete. Preskúmame niekoľko metód, ktoré produkujú vizuálne podobné výsledky ako súčasná, pomalá metóda Oklúzie. Aby sme zaistili kvalitu testovaných metód, nastavíme a vyhodnotíme 5 kvantitatívnych metrík …more
Abstract:
RationAI group trained a convolutional neural network model that can reliably predict the presence of prostate cancer in digitized tissue samples. Our goal is to find an explainability method to help us understand those predictions in a reasonable time. We review several popular explainability methods that produce visually similar results to the current, notably slow solution based on Occlusion. To …more
 

Keywords

digital histopathology convolutional neural networks explainable artificial intelligence occlusion CAM HiResCAM performance faithfulness localization
 
Language used: English
Date on which the thesis was submitted / produced: 23. 5. 2024
Identifier: https://is.muni.cz/th/l09kc/

Thesis defence

  • Date of defence: 24. 6. 2024
  • Supervisor: RNDr. Vít Musil, Ph.D.
  • Reader: Anselm Paulus

Citation record

Cite this text

Full text of thesis

Contents of on-line thesis archive
Published in Theses:
  • světu
Other ways of accessing the text
Institution archiving the thesis and making it accessible: Masarykova univerzita, Fakulta informatiky
Složka Reference to the local database directory of the institution