RF-DETR for Robust Mitotic Figure Detection: A MIDOG 2025 Track 1 Approach

Piotr Giedziun, Jan Sołtysik, Mateusz Górczany, Norbert Ropiak, Marcin Przymus, Piotr Krajewski, Jarosław Kwiecień, Artur Bartczak, Izabela Wasiak, Mateusz Maniewski · 2025

Read RF-DETR for Robust Mitotic Figure Detection: A MIDOG 2025 Track 1 Approach by Piotr Giedziun, Jan Sołtysik, Mateusz Górczany, Norbert Ropiak, Marcin Przymus, Piotr Krajewski, Jarosław Kwiecień, Artur Bartczak, Izabela Wasiak, Mateusz Maniewski (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-2509-02599.txt.

Description

Piotr Giedziun and nine co-authors present a 2025 computational pathology paper reporting their MIDOG 2025 challenge entry for spotting mitotic figures, dividing cells, in tissue slides. Scanners, stains, and tissue types vary widely, so models must generalise across domains.

Time limits forced the team to drop a planned two-stage design and refine a single-stage detector instead. They used RF-DETR, a transformer-based detector, with hard negative mining, which feeds the model its most confusing non-cell examples during training. Training drew on the MIDOG++ collection. When the preliminary test set was scored, the method posted a solid F1 score with recall ahead of precision. The authors stress that training data balance and hard negatives matter most for crossing domain shifts.

Catalog record

CREATORPiotr Giedziun, Jan Sołtysik, Mateusz Górczany, Norbert Ropiak, Marcin Przymus, Piotr Krajewski, Jarosław Kwiecień, Artur Bartczak, Izabela Wasiak, Mateusz Maniewski
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE11 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-02599

Browse the full library · About the Tower of Babel