YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
SENTRY-10 acoustic safety classifier
MIT/ast-finetuned-audioset-10-10-0.4593 fine-tuned on an 11-class subset of ESC-50 (10 safety-relevant classes + background_other), trained on folds 1-4 and evaluated on fold 5.
Classes: glass_breaking, siren, clock_alarm, crying_baby, chainsaw, hand_saw, car_horn, engine, vacuum_cleaner, footsteps, background_other. Confidence gate tau = 0.45.
artifacts/ holds the demo assets used by the SENTRY-10 live demo notebook.
Data credit: ESC-50 (K. J. Piczak, 2015), licensed CC BY-NC 3.0. The demo clips in artifacts/demo_audio/ev_*.wav are ESC-50 fold-5 clips; non-commercial use only.
Built as a 4-person team student project; not for safety-critical use.
- Downloads last month
- 25