About APR
Audio Pattern Ranger (APR) offers 24/7 monitoring for local disturbances in an environment, using machine learning models to detect and log specific nuisances, such as barking or car alarms. These models are trained on collected data to automate logging of detected disturbances.
Rather than using large complex solutions that make use of giant sample data, APR uses local recordings in order to identify the exact disturbance. This means that even an old laptop is plenty to put this project into action.
Background
In some jurisdictions, understaffing can lead to a lack of support for situations that are not life-threatening. In these cases, noise disturbances may be entirely ignored without an extended log of repeated violation along with video evidence proving log accuracy.
The primary purpose of this application is to simplify the collection and analysis of video footage to identify disturbances (e.g., dog barks) using a locally trained model. This model is designed to accurately detect and classify specific disturbances in the local area.
How It Works
Set up recorder
Collect some initial recordings
Extract individual noises
Train a model
Detect noises in collected recordings
Manually review the generated report
Refine model with additional training