Per-image results and trained models for: Recovering Silently Missed Small Lesions in Colonoscopy and Thyroid Ultrasound Segmentation Under Domain Shift

This record accompanies the manuscript "Recovering Silently Missed Small Lesions in Colonoscopy andThyroid Ultrasound Segmentation Under Domain Shift". It contains the per-image results from which everytable and figure of the manuscript is computed, the configuration and per-epoch learning curve of eachof the 144 training runs, the scripts that compute the tables and figures from these files, and thetrained weights of the models that carry the main results. The public datasets are not redistributed.The record holds two archives: `small_lesion_verification.zip` (results, run files, test index, andscripts) and `small_lesion_models.zip` (the `models/` folder with `predict.py`). Unpack both into thesame folder.