AutoWEEEdakt

Purpose and goal

The project´s overall objective was to develop an AI-based image recognition system that can identify and categorize electronic waste by observing how human process operators handle the waste. Manual work is used extensively in the recycling of electronic waste to pre-sort the waste before it is further processed. This is to ensure that electronic waste is recycled in a good and safe way. The purpose of this project was to investigate opportunities to increase the efficiency and safety of this work by utilizing and developing new AI-based methods.

Expected results and effects

AI-based methods were developed that automate the generating of annotated datasets based on monitoring manual work at two different electronics recycling facilities (El-Kretsen sorting station in Arboga and NG Metalls recycling facility in Katrineholm). These annotated datasets have there after been used to train AI models, which show good precision in categorizing electronic waste according to the categories which the waste should be sorted under. In the future, this could lead to a reduced need for manual work and increase safety at these facilities.

Planned approach and implementation

The project was divided into different work packages, where the first step was to review existing processes and make as small modifications as possible in order to not disturb existing processes but enable data collection. Next, the necessary hardware was installed to be able to collect data at the recycling facilities. This step was somewhat delayed due to the impact of the pandemic. Once data collection capabilities were in place, the focus was on developing algorithms and software that can automatically generate annotated data which AI-models could be trained and tested on.