Classifier

Overview

The classifier is a Python-based gRPC service responsible for real-time acoustic threat classification. It is designed to process audio frames and accurately identify various anomalies, specifically:

• UAV (Drones)
• Explosion
• Siren
• Truck
• Generator

ML Pipeline

This service utilizes a dual Machine Learning pipeline to ensure robust classification:

1. YAMNet: Used primarily for embeddings and feature extraction from the incoming audio frames.
2. Random Forest: Consumes the features extracted by YAMNet to perform the final classification into the defined threat categories.

Environment Variables

The service can be configured using the following environment variables:

Variable	Description	Default
CLASSIFIER_PORT	The port on which the gRPC service listens.	3232
LOG_LEVEL	Logging level for the application.	INFO
MODEL_PATH	Path to the Random Forest model file.	models/random_forest_v1.joblib
MODEL_VERSION	Identifier for the loaded model version.	rf-v2.0
YAMNET_MODEL_PATH	Path to the YAMNet TensorFlow Lite model.	models/yamnet.tflite
EXPLOSION_THRESHOLD_DB	Decibel threshold for detecting potential explosions.	-10.0
UAV_THRESHOLD_DB	Decibel threshold for detecting potential UAVs.	-30.0
YAMNET_EXPLOSION_CONFIDENCE	Minimum confidence score for explosion classification.	0.10
YAMNET_UAV_CONFIDENCE	Minimum confidence score for UAV classification.	0.10
YAMNET_SIREN_CONFIDENCE	Minimum confidence score for siren classification.	0.10
YAMNET_TRUCK_CONFIDENCE	Minimum confidence score for truck classification.	0.10
YAMNET_GENERATOR_CONFIDENCE	Minimum confidence score for generator classification.	0.10

Setup & Running

Local Development

1. Navigate to the classifier directory.
2. Install the required Python dependencies:

   pip install -r requirements.txt

3. If necessary, generate the gRPC stubs from your .proto files (assuming standard grpc_tools.protoc usage).
4. Run the service:

   python src/main.py

Running via Docker

The application is fully containerized. To run it using Docker:

docker build -t acoustic_guard_classifier .
docker run -d -p 3232:3232 --env-file .env acoustic_guard_classifier