AdaProj: Adaptively Scaled Angular Margin Subspace Projections for Anomalous Sound Detection with Auxiliary Classification Tasks
ASD system utilizing the AdaProj loss for acoustic machine condition monitoring for task 2 "First-Shot Unsupervised Anomalous Sound Detection for Machine Condition Monitoring" of the DCASE2023 Challenge.
Just start the main.py script for training and evaluation. To run the code, you need to download the development dataset, additional training dataset and the evaluation dataset, and store the files in an './eval_data' and a './dev_data' folder.
Please reference the following paper when reusing (parts of) the code:
@unpublished{wilkinghoff2024adaproj, author = {Wilkinghoff, Kevin}, title = {Ada{P}roj: {A}daptively Scaled Angular Margin Subspace Projections for Anomalous Sound Detection with Auxiliary Classification Tasks}, note = {arXiv:2403.14179}, year = {2024} }