Engineers build AI system using Raspberry Pi to sort waste
, by Sukanya Eiampinit, 1 min reading time
Engineers build AI system using Raspberry Pi to sort waste
Karl Myers and engineers from the Department of Mathematics and Computer Science at Liverpool Hope University have adapted a low-cost Raspberry Pi 3 computer to include a high-resolution camera to capture images for AI processing. Computer vision algorithms are used to precisely control a robot that sorts waste.
The UK will separate the recyclables for reuse and then send the rest to landfill. Therefore, the government is asking the public to help separate the recyclables into the same container before being collected and sorted at the waste separation and reuse plant. The waste will be sent along a conveyor belt and the staff will manually separate the recyclables. Tin and iron cans will be drawn out by magnets, while aluminium cans will flow along the conveyor belt into the container. However, the staff's care is error-prone, inefficient and expensive.
To that end, engineers used TrashNet, Google's dataset of 3,500 images of trash, to train a machine learning algorithm that could learn from one problem and apply it to new, related problems. The algorithm achieved 92% accuracy and 90% baseline efficiency, which could be put to commercial use.
The Raspberry Pi testbed takes images of waste on a conveyor belt in a waste sorting plant using a camera, processes them with an algorithm, and controls the robot to sort the waste accurately, at a cost of less than £100 (4,015 baht), but it's slower than human sorting. However, the technology's high accuracy and low cost more than offset the system's slow performance.
The research team is currently conducting experiments to further expand the capabilities of the algorithm and its efficiency. Using automated waste recycling technology could make the system more efficient by sorting waste at collection points instead of taking it to a landfill.
It is believed that using AI with Raspberry Pi will significantly reduce costs and manpower, and it is expected that in the future it will be increasingly applied in the industrial sector.