XPRIZE-Brazilian-Team/Bengt-BirdNET-Annotator · Apply for community grant: Academic project (gpu)

Hi,

I'm an artificial intelligence researcher focusing on developing interactive machine learning tools to support ecologists in handling large amounts of acoustic data.

I'm currently demonstrating the capabilities of Hugging Face Space, which allows users to annotate species-specific bird calls in audio datasets collected through passive acoustic monitoring. While the current functionality works, the speed of the inference process is relatively slow. Our team has established partnerships with ecologists responsible for managing ecosystem resources in some of the world's most biodiverse regions. Thankfully, I've been able to get feedback from some of them to help me refine the design through iterative improvements.

My overall goal is to create a tool that will benefit the ecology community by providing a general-purpose solution for AI-assisted labelling of large bioacoustic datasets. This tool could significantly contribute to the creation of large datasets with labels of the highest quality. To ensure a smooth user experience and avoid frustration, I plan to increase the speed of the application by incorporating GPU acceleration. The core of the application relies on a substantial pre-trained convolutional neural model implemented using TensorFlow.

Although I have access to significant High Performance Computing (HPC) resources within my institution, an important part of the impact of my research will come from creating solutions that are widely applicable. I believe that the Hugging Face platform offers a great opportunity to reach a wider audience.

To the best of my knowledge, this is the first application showcasing the state-of-the-art bird vocalization classifier BirdNET-Analyzer on Hugging Face.

Best,
Bengt