Training a BirdNet Object Classifier on Azure Machine Learning (AML)

Folder Overview

This folder contains information about how we train a custom BirdNet object classifier using Azure Machine Learning (AML). The training pipeline is designed to handle custom acoustic data, perform model training, and store the custom model on Azure Blob Storage.

Current Pipeline Details

• Pipeline Name: BirdNet-Custom-Training-Pipeline
• Datastore:
  • customaudiodata: Stores both the training data and the trained model.

Key Features

• Custom Model Training: The pipeline trains BirdNet on custom acoustic datasets.
• GPU Acceleration: Utilizes GPU clusters for efficient model training.
• Model Storage: Trained models are stored in Azure Blob Storage for further use.

Documentation

For detailed instructions, see the documentation files in the docs/ folder:

• Pipeline Architecture
• Usage Instructions
• Developer Guidelines