Page 210 - AI for Good Innovate for Impact
P. 210

AI for Good Innovate for Impact



                           o  Alcohol cans (e.g., beer cans)
                           o  US cans (imported beverage cans subject to distinct deposit regulations)
                           o  Bi-metal cans (composite cans combining steel and aluminum)

                           Fine-grained feature recognition, focusing on subtle differences in material surface,
                           texture, and labeling, was integrated to ensure high classification accuracy.
                      •    Rule-based Constraints:
                           Region-specific rules are also incorporated into WEE through a rule-based logic layer,
                           enforcing hard constraints where necessary.

                      Operational Feedback and Continuous Improvement

                      Feedback from pilot deployments, audits, and field operators is systematically collected and
                      used to refine WEE's performance:

                      •    Operational feedback identifies new waste types, emerging materials, and classification
                           challenges.
                      •    Audit results are analyzed to adjust detection thresholds or feature matching parameters.
                      •    Regulatory feedback is used to update classification rules as recycling standards evolve.

                      Through this dynamic feedback loop, WEE maintains strong alignment with country-specific
                      regulations while continuously improving operational robustness.

                      2�2     Benefits of use case

                      •    Prevent municipal waste from entering the environment through efficient treatment and
                           recycling, thereby reducing water and soil pollution and protecting ecosystems.
                      •    Improve urban waste management and air quality, reduce GHG emissions from
                           transport and treatment, and enhance infrastructure sustainability and eco-friendly urban
                           development.
                      •    Minimize waste generation and promote reuse through well-separated recyclables.
                           Optimize plastic waste sorting to increase recycled plastic usage and support circular
                           economy goals.
                      •    Reduce land-based marine pollution by using automated sorting systems to keep plastics
                           out of oceans, preserving marine ecosystems and enabling sustainable ocean use.
                      •    Higher recycling rates mean less landfill and incineration, reduced transport emissions,
                           and more circular resource flow. This benefits air, soil, and marine environments. The
                           Busan Facilities Corporation project proved that recycling is possible in previously
                           unsorted areas, resulting in higher recycling rates and better-quality secondary materials.


                      2�3     Future Work

                      We are moving to innovate AI-based recycling systems, improve rates, and secure high-
                      quality recycled raw materials. This requires finely tuned classification algorithms, robotic
                      implementation, and logistics optimization through data.

                      Our next step is classifying plastic compositions. As global regulations mandate recycled plastic
                      usage, we are developing high-speed plastic classification models using near-infrared sensors
                      and vision systems.

                      We continuously explore new sorting needs and develop models for finer classification. Our
                      team is optimizing model accuracy and size. Entering new markets requires localized data. We






                  174
   205   206   207   208   209   210   211   212   213   214   215