Are We Ready for Out-of-Distribution Detection in Digital Pathology?
Digital Pathology
DOI:
10.48550/arxiv.2407.13708
Publication Date:
2024-07-18
AUTHORS (3)
ABSTRACT
The detection of semantic and covariate out-of-distribution (OOD) examples is a critical yet overlooked challenge in digital pathology (DP). Recently, substantial insight methods on OOD were presented by the ML community, but how do they fare DP applications? To this end, we establish benchmark study, our highlights being: 1) adoption proper evaluation protocols, 2) comparison diverse detectors both single multi-model setting, 3) exploration into advanced settings like transfer learning (ImageNet vs. pre-training) choice architecture (CNNs transformers). Through comprehensive experiments, contribute new insights guidelines, paving way for future research discussion.
SUPPLEMENTAL MATERIAL
Coming soon ....
REFERENCES ()
CITATIONS ()
EXTERNAL LINKS
PlumX Metrics
RECOMMENDATIONS
FAIR ASSESSMENT
Coming soon ....
JUPYTER LAB
Coming soon ....