This project investigates the automated detection, classification, and retrieval of marine intertidal sounds from Passive Acoustic Monitoring (PAM) at urban and natural sites along the Eastern Cape coastline.

Using Computer Vision, Convolutional Neural Networks trained on spectrograms, the system distinguishes marine natural sounds (waves, soil,  air pressure and wave crashing), geophony (water flow). Integrated Explainable AI (saliency maps, SHAP) visualises decision drivers, enabling ecological validation.

A content‑based retrieval engine indexes spectrogram embeddings, allowing rapid similarity searches and efficient cataloguing of large acoustic datasets. The end‑to‑end pipeline provides scalable, real‑time monitoring to support biodiversity assessment, and tracking of human activities in coastal ecosystems.

This project creates an original radio drama addressing a social issue in South Africa. As the encoder, it will shape the sound design, dialogue, and performance to convey specific messages and emotions. The drama will be performed live to an audience, broadcast on Rhodes Music Radio, and available as a podcast.

​

This combines the immediacy of live performance with broader digital reach, aiming to spark conversation and social change. Drawing on Zakes Mda's work on storytelling's role in shaping public opinion, this project will test live radio's power to overcome traditional theatre's limitations and reach diverse audiences.

Coming soon