Titulo Estágio
AI-Powered Acoustic Monitoring of Biodiversity
Áreas de especialidade
Sistemas Inteligentes
Local do Estágio
CISUC
Enquadramento
The application of Artificial Intelligence (AI) to acoustic analysis is gaining momentum as a powerful tool for species monitoring and environmental assessment. In particular, the automatic recognition of bird vocalizations offers a non invasive, scalable, and efficient method for biodiversity tracking, complementing or even replacing traditional methods. Machine Learning (ML) models trained on audio data can detect species presence and activity across large spatial and temporal scales, contributing to conservation and ecological research.
This internship will focus on developing a sound recognition system for bird species classification using ML techniques. The project will explore Deep Learning (DL) architectures capable of processing audio signals, such as Convolutional Neural Networks (CNNs) applied to spectrograms and recurrent or transformer based models for sequential audio features. The initial dataset to be used is the British Birdsong Dataset available on Kaggle (https://www.kaggle.com/datasets/rtatman/british-birdsong-dataset), which includes bird calls and songs from various UK bird species.
However, this dataset is relatively small and unevenly distributed, and thus may not be sufficient for training high performing models. As such, the project will also consider strategies for data expansion, potentially through collaboration with biodiversity researchers or institutions that can provide access to larger and more diverse acoustic recordings. An important goal will be to adapt the model to recognize bird species relevant to local ecosystems, ensuring that the system is not only technically robust but also ecologically meaningful and regionally applicable.
Objetivo
The following objectives will guide the development of the audio-based classification system:
Study the state of the art in sound recognition using machine learning
Investigate suitable audio processing techniques and deep learning models (e.g., CNNs on spectrograms)
Analyze and preprocess the British Birdsong Dataset
Explore possibilities for expanding the dataset via partnerships with biodiversity research groups
Develop and evaluate a sound classification framework
Plano de Trabalhos - Semestre 1
Conduct a literature review on bird call recognition and machine learning methods for audio classification
Study preprocessing techniques (e.g., spectrogram generation, noise filtering)
Evaluate the quality and structure of the British Birdsong Dataset
Initiate contact with potential biodiversity partners to explore data-sharing opportunities
Design and begin implementing the audio classification framework
Write intermediate report
Plano de Trabalhos - Semestre 2
Finalize and optimize the model training and evaluation process
Integrate additional audio data if available
Test and evaluate performance
Write final report
Condições
This work should take place in the context of a research project in CISUC. There is the possibility of a 6-month scholarship.
Orientador
Catarina Silva, Bernardete Ribeiro, Dinis Costa
catarina@dei.uc.pt 📩