Seabin has launched an open-access data dashboard that tracks plastic pollution recovered from Sydney Harbour, including the brands, parent companies and packaging types identified among the litter.
Seabin's data dashboard is designed to give brands, packaging companies and the public greater visibility of plastic pollution entering the marine environment.
The platform draws on data collected through Seabin’s Sydney Harbour operations. Recovered material is counted, photographed and catalogued through its Ocean Health Lab before being added to the dashboard, which is updated daily.
Seabin CEO and co-founder Pete Ceglinski told PKN the platform reflects the organisation’s evolution from a clean-up technology provider into an environmental data business.
“We have a global data gap on what’s in the water [and] where it comes from,” Ceglinski said.
The dashboard allows Seabin to classify recovered material by plastic category, product, brand and parent company. Soft plastics are the second-largest plastic category identified in its data, behind microplastics.
Seabin has also started publishing rankings of the branded litter recovered from Sydney Harbour.
“We know that the brand is not the polluter, but the branded litter is in the water,” Ceglinski said.
He acknowledged that sales volumes and market share would affect how frequently a brand appeared in the data. Packaging without visible branding also cannot currently be attributed to a particular company.
Seabin is developing an AI system to accelerate the classification process, which is now carried out manually by its science and data team with support from citizen scientists.
The organisation has catalogued and tagged 87,000 items to help train the technology. Its first object-recognition test achieved 92 per cent accuracy, identifying 516 of 518 objects in an image.
Ceglinski stressed that this initial trial detected individual objects but did not identify the product or brand. Seabin’s development program will next focus on determining what each object is before progressing to brand recognition.
The system is intended to reduce the time spent manually identifying hundreds of thousands of items while retaining human verification of the results.
Ceglinski said access to the underlying images would also allow Seabin to substantiate its findings if the data were challenged.
The dashboard initially presents historical data through charts and other insights. Seabin plans to expand its capabilities to include further litter categories and, eventually, predictive modelling based on factors such as seasonal conditions, rainfall, population and location.
The dashboard can be accessed through Seabin’s website at seabin.io.
