acoupi: An Open-Source Python Framework for Deploying Bioacoustic AI on Edge Devices

Passive acoustic monitoring (PAM) has become an essential tool in biodiversity research, but traditional PAM systems have significant limitations. These systems often require manual data offloading and impose substantial demands on storage and computing infrastructure. To address these challenges, a team of researchers led by Aude Vuilliomenet has developed acoupi, an open-source Python framework that simplifies the creation and deployment of smart bioacoustic devices.

What Changed?

On-Device AI Processing:

• Technical Change: acoupi integrates on-device AI-based processing with network connectivity.
• Why It Matters: This combination enables local data analysis and transmission of only relevant information, significantly reducing storage needs. For researchers and conservationists, this means more efficient and scalable monitoring systems.

Modular Framework:

• Technical Change: acoupi modularizes key elements of the bioacoustic monitoring workflow.
• Why It Matters: Users can easily customize, extend, or select specific components to fit their unique monitoring needs. This flexibility is crucial for deploying AI models on custom hardware and tailoring device behavior.

Key Features

• Audio Recording:
  • Captures audio data from the environment using built-in microphones or external devices.
• AI-Based Data Processing:
  • Integrates pre-trained bioacoustic classifiers like BirdNET (for bird species) and BatDetect2 (for UK bat species).
• Data Management:
  • Handles data storage, compression, and transmission to cloud services or local servers.
• Real-Time Wireless Messaging:
  • Facilitates real-time communication between devices and a central server for monitoring and control.

Implementation Details

Hardware Compatibility:

• acoupi can be deployed on low-cost hardware such as the Raspberry Pi. This makes it accessible for researchers with limited budgets.
• The framework is designed to run efficiently on resource-constrained edge devices, ensuring that AI models can process data in real-time without significant latency.

Modularity and Extensibility:

• Each component of acoupi (audio recording, data processing, etc.) is modular. This allows users to swap out or add new components as needed.
• The framework supports a plugin architecture, making it easy to integrate custom AI models or third-party libraries.

Real-World Deployment

To demonstrate the reliability and flexibility of acoupi, the researchers conducted a month-long deployment in a UK urban park using two acoupi-powered devices. Here are some key findings:

• Reliability:
  • The devices operated continuously for the entire duration without significant issues.
• Data Quality:
  • Both BirdNET and BatDetect2 classifiers performed well, accurately identifying bird and bat species in real-time.
• Storage Efficiency:
  • By transmitting only relevant data, the system significantly reduced storage requirements.

Benefits for Practitioners

Ease of Use:

• acoupi's user-friendly interface and documentation make it accessible to researchers with varying levels of technical expertise.
• The framework abstracts away many of the complexities associated with embedded systems programming, allowing users to focus on their monitoring goals.

Scalability:

• The modular design allows for easy scaling. Users can deploy multiple devices in different locations and manage them centrally.
• acoupi's support for real-time wireless messaging enables centralized monitoring and control, making it ideal for large-scale biodiversity projects.

Conclusion

acoupi represents a significant step forward in the deployment of AI-powered PAM systems. By providing a unified and configurable framework, acoupi simplifies the process of creating and deploying smart bioacoustic devices. This tool has the potential to revolutionize how researchers and conservationists monitor biodiversity, making it more efficient, scalable, and accessible.