robustness-of-transferability-estimation-metrics-for-medical-imaging

This repository provides the weights of the fine-tuned models used to obtain the reference rankings in the work mentioned below.

File structure:
├── full                  <- fine-tuned on the full target datasets
│   ├── accuracy          <- optimized for accuracy
│   ├── auroc             <- optimized for AUROC
├── 5pct_seed42           <- fine-tuned on 5% fraction sampled with seed 42 
│   ├── accuracy          
│   ├── auroc             
└── 5pct_seed43           <- fine-tuned on 5% fraction sampled with seed 43
    ├── accuracy          
    └──  auroc          

Please cite the following paper if you end up using some of the models:

@misc{claßen2026robustnesstransferabilityestimationmetrics,
      title={Robustness of transferability estimation metrics for medical imaging}, 
      author={Niclas Claßen and Théo Sourget and Dovile Juodelyte and Rob van der Goot and Veronika Cheplygina},
      year={2026},
      eprint={2608.09999},
      archivePrefix={arXiv},
      primaryClass={eess.IV},
      url={https://arxiv.org/abs/2608.09999}, 
}