Data and code¶
Multimodal interaction analysis¶
Analysing structure and patterns in multimodal interactions is challenging because signals often overlap. The Multi-Interaction tool supports analyses based on behavioural annotations from BORIS or ELAN. Implementation details and motivation are described in the associated publication.
Marmaudio¶
Marmosets are a useful model in neuroscience and voice-perception research. A large-scale dataset of marmoset vocalisations is available with an associated article.
Fundamental frequency estimation¶
Extracting frequency contours is a common challenge in bioacoustic analysis.
- Use the Python interface to estimate fundamental frequency and detect tonal signals.
- Evaluate estimators with the cross-species benchmark dataset.
- Read the associated publication.
Cetacean vocalisation detection¶
Ceta-CNNs is a suite of pretrained convolutional neural networks for detecting cetacean vocalisations in passive recordings. Covered species include orcas, humpback whales, delphinid whistles, fin whales, sperm whales, and blue whales.
Vocalisation clustering based on deep embeddings¶
Animal vocalisations typically fall into discrete categories, including individual signatures and call types.
- Use the repertoire embedder to cluster vocalisations.
- Evaluate models with the cross-species benchmark dataset.
- Read the associated publication.