AnoPatch: Towards Better Consistency in Machine Anomalous Sound Detection

TL;DR

AnoPatch introduces a patch-based fine-tuning approach using a pre-trained ViT model on AudioSet to improve consistency and performance in machine anomalous sound detection, achieving state-of-the-art results.

Contribution

The paper proposes AnoPatch, a novel patch-level fine-tuning method with a ViT backbone pre-trained on AudioSet, tailored for machine audio anomaly detection.

Findings

Achieves state-of-the-art performance on DCASE 2020 and 2023 ASD datasets.

Demonstrates that better model-data consistency improves detection accuracy.

Empirically shows the effectiveness of patch-level modeling for machine sounds.

Abstract

Large pre-trained models have demonstrated dominant performances in multiple areas, where the consistency between pre-training and fine-tuning is the key to success. However, few works reported satisfactory results of pre-trained models for the machine anomalous sound detection (ASD) task. This may be caused by the inconsistency of the pre-trained model and the inductive bias of machine audio, resulting in inconsistency in data and architecture. Thus, we propose AnoPatch which utilizes a ViT backbone pre-trained on AudioSet and fine-tunes it on machine audio. It is believed that machine audio is more related to audio datasets than speech datasets, and modeling it from patch level suits the sparsity of machine audio. As a result, AnoPatch showcases state-of-the-art (SOTA) performances on the DCASE 2020 ASD dataset and the DCASE 2023 ASD dataset. We also compare multiple pre-trained models and empirically demonstrate that better consistency yields considerable improvement.