Using photos from the public to monitor river pollution

The Water Network

In what may be a world first, a community group in Bradford, UK, has gathered photographs to help train an AI model to identify visual markers of river health.

The Water Research Centre led the work, with funding from Natural England, and in collaboration with Rain++ and students at the National Taiwan University. This involved using 15,700 photos to train an AI model to identify visual markers of river health, including rubbish, wildlife and discoloration.

There are hopes that the model could be used more widely across the UK. The model and documentation have been made available on GitHub, although are still under development.

Read more about this project here.