“With great power comes great responsibility.” — Uncle Ben, Spiderman
Who knew a comic book character could hold such wisdom? Yet, this maxim stays true even in reality. Although humanity’s latest innovation, electronics, has drastically improved our quality of life, it’s not dealt with responsibly.
Every single year, 50 million metric tons of electronic waste is thrown out, and only one-eighth of that is being properly recycled. The rest?
They sit in landfills gathering rust, destroying our planet, and harming people all at the same time.
Then why doesn’t someone do something about this?
Recycling e-waste is unpopular because it’s too inconvenient. Instead of just throwing it in the trash, consumers have to go out of their way to recycle.
This problem has been overlooked for far too long. The mountains of dilapidated computers, phones, and other consumer electronics have grown too massive to simply sweep under the rug.
This issue facing humanity requires a fix that is both efficient and cost-effective. By applying the rapidly advancing field of artificial intelligence with the global e-waste crisis, our long-term solution provides a sustainable system to recycle e-waste.
What is E-Waste?
E-waste is defined as any electronic equipment that’s been discarded, including home appliances (microwaves, fridges), consumer electronics (phones, computers), and electronic utilities (light bulbs, batteries).
Most electronics contain many toxic materials such as lead, cadmium, lithium, barium, mercury, and arsenic that can contaminate the soil, water, and air. This can cause significant harm to humans as well as wildlife.
If the toxic chemicals contained in e-waste contaminate local water sources (rivers, groundwater) or are released into the air, they can harm people. The health risks range from kidney disease and brain damage to genetic mutations and cancer.
Current “Solutions”
Most e-waste is not recycled. Instead, it is thrown in large pits or burned. Needless to say, both of those options are terrible for the environment as it releases toxic chemicals into the soil, waterways, and air.
Some recyclers ship collected e-waste to underdeveloped areas of the world under the guise of philanthropy. Claiming that their “donations” help bring technology to developing nations, they illegally dump the trash in inhabited areas.
One of these places is the city of Guiyu in rural China. Due to e-waste contaminating the drinking water there, it has the highest level of cancer-causing dioxins in the world. Scientists also found that 82% of the children living there had lead poisoning.
For the small percentage of e-waste that is recycled, the process is extremely inefficient. Though electronics often contain a variety of useful and expensive minerals and metals such as gold and platinum, most processes only salvage the steel, copper, and plastic due to the outdated technology employed.
E-Cycle
E-Cycle is our proposed solution to this global crisis. It utilizes image processing AI to sort e-waste more efficiently. The process consists of 3 main steps: collection, recognition, and separation.
Collection
First, all of the e-waste is collected at a central location. This can include anything from discarded smartphones and printers to old wires and batteries.
The variety of e-waste is then placed on a conveyor belt that passes them through the AI image recognition system.
Recognition
The image processing algorithm to be implemented is called SSD (single shot multibox detector). This algorithm is most optimal for this application due to its ability to identify objects with reasonable accuracy in short amounts of time.
It’s able to achieve such a quick turnaround time because the object classification is done in a single forward path of the network, hence the name “single shot.” “Multibox” refers to the technique of bounding box localization.
The AI model is comprised of a series of three convolutional neural networks: the base network, the auxiliary convolutions, and the prediction convolutions.
The base convolutions are derived from existing image classification architecture that is pre-trained on open datasets like OpenImages as well as our own datasets. The auxiliary convolutions are added on top of the base to provide higher-level feature maps.
A feature map is simply an output of one filter that is applied to the raw image or a previous layer.
Finally, the prediction convolutions locate and identify the objects in the feature maps.
Initially, our AI model will be trained on open datasets such as OpenImages. But as time passes, our model will be able to achieve higher and higher levels of accuracy due to the immense amount of data collected throughout the image recognition process that can be used to further train it.
Separation
After the AI has classified the types of e-waste approaching on the conveyor belt, it relays the information to the robotic tube. This tube harnesses the force of vacuum suction to be able to better grab objects of different shapes and sizes.
The arm will position itself over the object on the conveyor belt that the AI has classified and located. The object will then be trapped against the mesh lattice covering the tip of the tube as a result of the strong inflow of air up through the tube.
Then, the air is able to move and drop the object into one of the multiple labeled areas for the different types of e-waste and metals.
The impact of E-Cycle
The global e-waste problem is monstrous. If we can recycle even 1% of our annual e-waste production, that would improve the quality of life for millions.
More specifically, 1% less e-waste means 2 million fewer lives put at risk.
Tapping into the rare commodities market with 1% of the e-waste material means $6.25 billion in potential revenue.
Moreover, our process is projected to be much more impactful than recycling 1% of e-waste. With the model having an average accuracy of 46.4%, we could be impacting many more people and generating much more revenue.
The ripple effects are boundless. Implementing E-Cycle would not only conserve energy and raw materials, but it would also save and improve the lives of people as well as wildlife.
Created by: Jason Liu, Ethan Wei, Nisha Lerdsuwanrut, Katelyn Won, Nikitha Ambatipudi, & Max Holschneider
