<![CDATA[The rapid accumulation of inorganic waste in urban environments requires an effective and sustainable technological approach. Conventional sorting methods that rely entirely on human awareness are often inefficient, labor-intensive, and prone to error, leading to minimal reduction in waste volume at the source. This study aims to design an automatic waste sorting system named ”Orange Box” utilizing Artificial Intelligence and Computer Vision methods implemented on a Raspberry Pi 5 to achieve high-speed and accurate waste classification. The system employs a V2 Camera Module for visual data acquisition and the BCM2712 processor to run a MobileNetV2 model trained on a custom dataset of 50 images. The waste is classified into plastic bottle and cup. Physical separation is executed by MG995R Servo Motors supported by a 20A Power Supply Unit.Performance testing demonstrates that the system achieves a classification accuracy of 92% with an average inference time of 45 Ms per image. Electrical analysis shows stable operation with an average power consumption of 10.50 – 11.10 Watt during processing and a peak load of 17.17 Watt during actuator movement. The integration of Raspberry Pi 5 successfully overcomes latency issues found in previous studies, providing a real-time, energy-efficient, and consistent sorting solution suitable for deployment in public facilities and smart city infrastructures.]]>
Copyrights © 2026
