We’re polluting our planet with our trash.
Everything we “throw away” ends up somewhere else — destined for landfill or even the oceans. With increasing awareness of environmental calamities like mountains of trash in some Asian countries and the Great Pacific Garbage Patches, out of sight is no longer necessarily out of mind.
Waste is part of the everyday lives of consumers, so it can be hard to see what we can do to help. Recycling can help tackle the problem, but even this isn’t as simple as it sounds.
Bin-e is a Poland-based startup and member of the NVIDIA Inception program that wants to take the guesswork, and some of the guilt, out of our relationship with waste. The company uses image recognition algorithms powered by NVIDIA Jetson TX1, a supercomputer on a module for AI at the edge applications.
Some local governments sort waste for you. Others ask its citizens to diligently separate different types of plastics, metal, glass and other materials.
In the workplace, recycling can get even more difficult, with many offices offering very limited recycling facilities or none at all.
Bin-e’s smart bin system makes recycling a snap. Simply pop your trash inside and it automatically separates it, making it ready for recycling.
The Dirty Truth About Recycling
A crucial reason for this low recycling rate is inadequate sorting of materials. If different types of trash are mixed together in the batches sent to recycling centers, it can mean that the whole lot cannot be processed correctly and winds up in landfills.
“In 2015, only 19 percent of waste from EU capitals was collected separately, meaning that 80 percent still ended up in residual bins,” explained Jakub Luboński, co-founder and CEO of Bin-e. “In office and public spaces, it remains difficult to introduce an efficient sorting system and recycling guidelines are often unclear.”
To meet European Union recycling targets by 2025, municipalities need to increase their recycling rates to 55 percent. This is where Bin-e’s smart bin comes into play.
The startup’s multifunctional bin recognizes, sorts and compresses waste automatically. It can be trained to identify a number of different waste types — from paper and plastic to aluminum and e-waste.
When trash is thrown into the bin, a camera captures an image of it, which is then processed by an embedded Jetson TX1 module. Using a series of AI algorithms, the module quickly identifies which of the configured categories the trash belongs to. Once this is determined, the trash item is deposited in the correct bin.
“When we were developing our system, we knew we needed to utilize hardware which could provide the computing power necessary to identify and sort an infinite number of waste items,” said Luboński. “We chose to work with Jetson TX1 as it is very small, very powerful and extremely cost effective — for us, there was no alternative.”
Currently, Bin-e operates at an accuracy level of 80 percent. To achieve this, the company’s developers trained a series of deep learning algorithms with a set of image recognition scripts on NVIDIA TITAN Xp GPUs running TensorFlow and Keras.
Not Just Good for the Planet
Not only is Bin-e’s offering good news for the environment, but also for companies. Thanks to smarter sorting, businesses need to spend less time and money on waste management and removal. In Poland, Bin-e is already used by companies such as McDonald’s and Harper Hygienics.
And because it’s trainable, Bin-e’s system can go beyond simply paper vs. plastic. It could boost security in airports or transit hubs by identifying suspicious-looking objects dropped inside it. And then automatically notify relevant teams for verification.