SuperBin, an AI Company in Pursuit of a Circular Economy

SuperBin is an AI company that pursues a circular economy through the selective collection and recycling of waste based on the phrase, “Trash is money, and recycling is a game.” SuperBin provides a service that allows individuals to receive points that can be converted into money whenever they discard recyclable resources such as PET plastics and cans through the use of an AI recyclable resource recovery robot and a platform that selectively collects these recyclable resources and then provides selected recyclable resources to processing companies to help produce high-quality recycled raw materials. In addition, the company also holds various recycling cultural projects using spaces, such as classes, exhibitions, shopping, hands-on experiences and games.

The AI Required to Increase the Value of Recycling

Clean transparent PET plastic (polyethylene terephthalate: a synthetic resin used in the manufacture of drinks bottles, etc.) is a recyclable resource that can be recycled to create high additional value as a high-quality raw material. However, during the waste collection process, due to the high quantity of foreign materials mixed in with the PET plastic most sorting being done manually, less than half of PET plastics are able to be recycled into a high-quality raw material.

Aside from this, more than a third of recyclable resources are not recycled and instead incinerated or put in landfill due to operational issues such as excess amounts of waste and limitations on the number of workers in the sorting plant.

SuperBin is building a new type of recycling process that selects recyclable resources through the use of AI with the aim of improving inefficiency in the recycling process and increasing recycling value. In order to simplify the recycling process, SuperBin is developing AI that can automatically sort recyclable resources found during the recycling process beyond the scope of sorted collection carried out by the AI recyclable resource recovery robot.

The Data Processing Required to Sort Recyclable Resources During the Recycling Process

The previous AI recyclable resource recovery robots were trained by SuperBin to process data using the bounding box method.

However, as waste to be sorted is becoming more diverse and often has foreign materials mixed in, the performance of AI had to be further advanced in order to be able to make more accurate distinctions. In order to do this, SuperBin processed additional data on the types and colors of recyclable resources, as well as foreign materials such as lids, labels and other contaminants using polygon segmentation, a method which processes along the outlines of objects.

Extensive Data Processing Experience and Professional Tools Provided by Testworks

SuperBin may have been able to commercialize AI that can sort and collect recyclable resources by processing waste data collected directly by using bounding boxes, but lacked the resources to perform polygon segmentation processing directly for their new recycling process. Since polygon segmentation processing is far more difficult than bounding box processing, it takes a lot longer to complete the process and professional processing tools are essential for processing as it is a task that requires high precision. Owing to this, SuperBin chose to partner with Testworks, a company that has professional manpower with extensive data processing experience and blackolive, which is an automated data processing tool. Thanks to the experience and expertise of Testworks, a data voucher supplier for the past four consecutive years, SuperBin was able to proceed with their first project with Testworks, and was selected by the government to run the data voucher support project with government support.

Data Processing That Highlighted The Expertise of Testworks

To process data according to the needs of SuperBin, Testworks kept in active communication with SuperBin in the initial stages of the project to define the data to be processed. Guidelines were established and checks were made for any issues that may arise during processing.

Based on established guidelines, Testworks’ data specialists used blackolive to process polygon segmentation and assured quality through systematic inspections. In particular, a four-step processing process was performed for more accurate processing, starting with primary processing being conducted by Testworks’ automation model, followed by adjustment and reprocessing by professional personnel, full inspection, and finally sampling inspection. In addition, polygon segmentation data was converted into a bounding box format dataset and provided in order to measure improvement effects by comparing performance of AI trained with previously built data in the output delivery stage.

“As the segmentation data we needed for sorting recyclable resources had a lot of different categories, it was very difficult to process them. However, we were able to successfully build data thanks to Testworks’ experienced professionals and the automation solution, blackolive. Being a social venture ourselves, we wish Testworks continued growth and success in their pursuit of social innovation based on technological prowess.”

Jung-bin Kim, SuperBin CEO

Putting the Much-Anticipated Circular Economy Into Practice Through a Successful Data Voucher Support Project

SuperBin and Testworks collaborated through a data voucher support project to process more than 157,000 objects, which was the target amount during the data construction period of roughly three months. Through this project, SuperBin was able to improve the AI of their existing recyclable resource recovery robot and obtained the necessary datasets to develop AI that can be introduced onto sorting lines in high-quality recycled raw material processing plants. While working on the project with SuperBin, Testworks was able to build an automated processing model for data on wastes and generated customer satisfaction using high-quality data.

Starting with the success of the data voucher support project, both companies aim to continue to build data on wastes for recycling processes using AI. Both aim to create a circular economy through innovation of the recycling process by building waste data and expanding the types of recognizable recyclable resources and developing solutions to recognizing and improving the treatment of all waste, including materials that are intended for incineration or landfill.