FrOoDo: Framework for Out-of-Distribution Detection

TL;DR

FrOoDo introduces a modular, extensible framework for Out-of-Distribution detection in digital pathology that unifies post-hoc OoD methods with flexible data generation strategies. It supports PyTorch classification and segmentation models and emphasizes reproducibility through a fixed architecture and identical image sampling across methods. The framework provides pathology-specific augmentations and dataset-wide OoD design options, plus implementations of standard OoD scoring methods such as the Max Softmax baseline, Energy-based scoring, and ODIN perturbations. By offering clear interfaces for datasets, methods, and augmentations, FrOoDo enables rapid, fair evaluation of new datasets, methods, or models, with documentation and code available at https://github.com/MECLabTUDA/FrOoDo.

Abstract

FrOoDo is an easy-to-use and flexible framework for Out-of-Distribution detection tasks in digital pathology. It can be used with PyTorch classification and segmentation models, and its modular design allows for easy extension. The goal is to automate the task of OoD Evaluation such that research can focus on the main goal of either designing new models, new methods or evaluating a new dataset. The code can be found at https://github.com/MECLabTUDA/FrOoDo.

Figures (2)

• Figure 1: Out-of-Distribution Evaluation Task Design
• Figure 2: Currently supported augmentations applied to an image of the BCSS dataset amgad2019structured and the resulting OoD images and masks