Funded project, 2024 faunanet

This project aimed at creating a software prototype that allows the deployment of arbitrary bioacoustic machine learning models on edge devices like the Raspberry Pi Zero, with a focus on maintainability and extendability. This prototype should eventually be deployed on a network of devices which can be accessed remotely. Main requirement was the ability to exchange the deployed models on the fly through a simple user interface. Existing solutions like birdnet-Pi are bound to a specific model and require model files to be exchanged by hand, which was not feasible in the proposed use case. 

• Software design: Choose a suitable basis for the system and design a suitable architecture that allows dynamic model exchange, incorporating client feedback
• Implementation: Implement a new software package capable of running on a Raspberry Pi Zero, adhering to best practices. 

• Redesign of existing solutions to allow for the desired flexibility. This has been accomplished by including an abstraction layer for data preprocessing and model execution.
• A cross-platform, fully tested Python-based package for deployment of arbitrary machine learning models has been created from scratch called faunanet.
• The package has been published as a open source software.