This directory contains evaluation scripts and utilities for analyzing model performance, comparing different approaches, and generating publication-quality figures.
The evaluation system supports two main workflows:
-
Automatic evaluation during training/inference: The
evaluate.pyscript runs automatically after training or when loading pre-trained models, generating metrics and visualizations. -
Post-hoc analysis and comparison: The standalone evaluation scripts in this directory (
evaluate_*.py) allow you to compare multiple models and generate comparative analyses. It assumes that you already have the samples (i.e., theevaluate.pyscript has already been run). The paths are coded for our specific runs, so you might need to modify them and adapt them to your own needs.
- See INFERENCE.md for details on running inference
- See TRAIN.md for training configurations
- Check the main README.md for overall project overview