The award winners used sensor data and machine learning (ML) at the edge to detect early signs of harmful algae blooms and notify residents about local water conditions.

Harmful algae blooms—a phenomenon caused when some types of algae that produce toxins grow out of control—can wreak havoc on aquatic ecosystems and harm people and pets.

The inventors, Sashrika Das and her father, Mithum Das, built their device using a Wio Terminal, a microcontroller-based, fast prototyping tool, equipping it with multiple sensors to detect the presence of algae in water. Designing an alert system involves capturing and comparing many types of sensor readings. They used the Edge Impulse development kit to design a TinyML model, enabling machine learning at the edge.

To train the model to predict water conditions based on captured sensor values, they collected a variety of water samples as an initial dataset—with plans to add to it in the future. The device took readings hourly from TDS (total dissolved solids), turbidity and pH sensors. They use the TinyML model to classify the algae level as normal, warning or dangerous. The data transmits to the AWS cloud over the Helium Network using a LoRa module connected to the Wio Terminal. A RESTful API was also developed to allow third parties, such as town halls and public health agencies, to subscribe to the data.

As part of the contest’s grand prize offering, Seeed Studio is working with the creators to prototype and test a version of their solution using the Wio WM1110 Dev Kit with the intention of adding it to the Seeed Co-Create Licensing Program, potentially making it available through Seeed's official store.