AI speeds object recognition for recycling - Embedded.com

AI speeds object recognition for recycling

Researchers from Liverpool Hope University in the UK are using artificial intelligence (AI) for sorting recycling in household rubbish.

The system developed by the team uses a Raspberry Pi running an AI object detection algorithm and a high-resolution camera to identify paper, cardboard, plastic, metal, and glass in household rubbish.

The algorithm uses transfer learning, a machine learning technique that takes a neural network used for one task and applies it to another domain, usually via additional training with a different dataset. In this case, the algorithm was trained on a database of 3500 images of rubbish from Google images as well as an online database called “TrashNet”. The resulting system was 92% accurate in tests.


Sorting recycling is currently done with both automatic and manual methods in the UK. The researchers hope to eliminate manual processes in the future (Image: Liverpool Hope University)

The system is designed to be a low-cost method for sorting recycling, with the prototype costing under £100 to build. The researchers aim is to keep cost low to enable the system to be adopted around the world, however, in its current form, it is too slow to be commercially viable. A commercial version would require more processing power than the Raspberry Pi could provide.

In their report, researchers Emanuele Lindo Secco and Karl Myers write: “Due to rapid urbanization, increasing population and industrialization, there has been a sharp rise in solid waste pollution across the globe. Do we have the capacity to handle such an increase in waste? The answer to the question is no – at present we do not have the capacity to handle an increase in waste and moreover we do not have the capacity to handle the waste we are retrieving now. Hence, efforts must be made to streamline the waste sorting process and also efforts into intelligent retrieval of waste to further ease the pressure on the MRFs (Material Recovery Facilities) must be increased.”


The algorithm was trained on various datasets for sorting recycling, like the plastic water and milk bottles seen here (Image: Liverpool Hope University)

Currently in the UK, both automatic and manual methods are used for sorting recycling. As the rubbish passes by on a conveyor belt, strong magnets are used to separate steel and cans, with a “reverse magnet” producing eddy currents in aluminium cans that pushes them from the belt. Humans have to manually pick paper and plastic from the belt, as well as sorting metal items the previous step missed. The idea behind this new research is to replace the humans with robots, equipped with this visual AI system that can identify the types of rubbish as the conveyor goes past. Robot arms can then pick out and sort the different types of rubbish.

>> This article was originally published on our sister site, EE Times Europe.

 


For more Embedded, subscribe to Embedded’s weekly email newsletter.

 

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.