Delivering ROI and helping the environment using Image Recognition
In 2018, a major metal recycling company that handle and recycle 1.8 million tonnes of material per year approached us about helping them identify stockpiles of cars and other metal waste on a national scale. At the time, they used manual methods to identify stockpiles at great expense to the company. They were looking to modernise this approach, speed up stockpile identification, increase the frequency of stockpile identification and reduce costs.
Over the course of our close collaboration with our partner’s technical team, it became clear that we could deploy Remi AI’s image recognition across satellite and aerial imagery to identify stockpiles.
We set about integrating our image recognition platform into
their own software systems and cloud infrastructure. Although our image recognition was already capable of identifying cars from a topographical perspective, we had to test the proximity thresholds of cars to each other in order to not falsely classify car parks as stockpiles of disused cars.
Within two weeks, we were able to deploy Remi AI’s image recognition platform to a new cloud infrastructure and run it to identify over 59 previously unidentified stockpiles in Western Australia alone. This was later rolled out on a national basis. Although the project is still ongoing, the estimated ROI on this project is 380%.