#tfgram
One‑stage supervised contrastive learning (OS‑SCL) reached 94.64% AUC on DCASE 2020, rising to 95.71% with the new TFgram feature. Code is available on GitHub. Read more: https://getnews.me/supervised-contrastive-learning-improves-anomalous-sound-detection/ #anomaloussounddetection #tfgram
September 20, 2025 at 3:05 PM
One-stage supervised contrastive learning (OS-SCL) with feature perturbation addresses false alarms in anomalous sound detection; TFgram feature improves performance on DCASE 2020.
Noise Supervised Contrastive Learning and Feature-Perturbed for Anomalous Sound Detection
Shun Huang, Zhihua Fang, Liang He
arxiv.org
September 18, 2025 at 9:31 AM