F1-EV Score: Measuring the Likelihood of Estimating a Good Decision Threshold for Semi-Supervised Anomaly Detection
Implementation of the threshold-independent performance measure F1-EV and its bounded version for semi-supervised anomaly detection. The script is designed to be evaluated with output files of the anomaly detection tasks of the DCASE Challenge.
Just run the script and pass the folders containing the predictions and the folder containing the labels as arguments: python f1_ev.py -pred_files_path ./dcase-2023/teams/ -ref_files_path ./dcase-2023/ground_truth_data/ -alpha_test 0
When reusing (parts of) the code, a reference to the following paper would be appreciated:
@inproceedings{wilkinghoff2024f1-ev, author = {Wilkinghoff, Kevin and Imoto, Keisuke}, title = {{F1-EV} Score: Measuring the Likelihood of Estimating a Good Decision Threshold for Semi-Supervised Anomaly Detection}, booktitle = {International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, year = {2024}, publisher={IEEE}, pages={256--260} }