GIC is continuing its efforts to fight plastic litter in Southeast Asia by launching a new citizen science tool called pLitter on World Environment Day this Saturday June 05, 2021.

The launch will take place online from 09:00-10:00 (Bangkok time), along with helpdesk support available from 15:00-16:00. Helpdesk support will continue at the aforementioned times for the days following World Environment Day through June 08, 2021.

See the attached flyer for details about the World Environment Day event.

What is pLitter?

pLitter is an online image annotation platform developed by GIC to support the United Nations Environment Program’s (UNEP) CounterMEASURE project. pLitter gives citizen scientists the power to improve their communities’ environmental health by training a machine learning model to automatically identify plastic litter. The model is still in its early stages, similar to a toddler, so its counting on citizen scientists to teach it what plastic litter looks like.

The Power of Citizen Science

Machine learning algorithms require comprehensive, representative training datasets in order to make accurate predictions. Providing such a dataset could be a daunting task if left to a small team. Citizen Scientist involvement can reduce annotation times from weeks to mere days. By engaging environmentally conscious citizen scientists from around Southeast Asia, we aim to create a thorough training dataset to strengthen GIC’s plastic litter identification model.

How to use pLitter

Users who access pLitter online will mark and classify, or annotate, plastic litter that appears in roadside scenes taken by vehicle-mounted cameras. In a typical annotation, users will draw a bounding box around plastic litter in the image, then assign a class to it. There are numerous plastic classes available in the platform including bottles, bags, and food containers to name a few, as well as piles of plastic litter, facemasks, and rubbish bins. The pLitter interface was designed to be simple, yet intuitive so that new users can confidently perform plastic litter annotations.

Equally as important to this process is teaching the model what is not plastic litter. Occasionally users will encounter errant bounding boxes from the model’s plastic predictions that are not plastic litter. These boxes should be flagged or deleted by users to improve the model.

The Way Forward

After sufficient training from citizen scientists with pLitter, the machine learning model will be able to automatically identify plastic litter with a high degree of accuracy. The GIC team will then use the output from the machine learning model to create plastic litter hotspot maps that will be shared with UNEP and relevant policymakers to forge a more sustainable future for the region.

Click here to access the pLitter online platform and start training the plastic litter identification model. 

Following the launch on June 05, the pLitter image annotation campaign is set to run through March 2022.

pLitter is part of a collaboration between GIC, UNEP, and Google.