<![CDATA[This study aims to develop a waste detection and classification application using the SSD MobileNetV2 algorithm based on an Android application. The problem of waste, especially waste generation, is a crucial issue that needs to be addressed, one of the factors being the lack of public awareness regarding waste sorting. Various efforts to increase public awareness of waste sorting that have been carried out, such as socialization, counseling, the use of brochure media, and poster media, require a lot of time, effort, and resources. This research was conducted to propose a more efficient and practical approach to education as well as handling waste sorting, namely by using an Android application integrated with the SSD MobileNetV2 object detection algorithm. The method used in this study consists of the process of collecting datasets of waste objects with types of organic, inorganic, and hazardous and toxic materials (B3), then training the SSD MobileNetV2 algorithm using the MediaPipe framework with the mediapipe-model-maker library, and developing an Android application integrated with the trained SSD MobileNetV2 algorithm using the MediaPipe framework with the Mediapipe Tasks Vision library. This study produced a synthetic dataset in Pascal VOC format with a total of 4302 images of waste objects divided into 80% for the training set and 20% for the validation set. The created dataset was then trained on the SSD MobileNetV2 algorithm with performance results of AP IoU=0.50:0.95 with a value of 0.847, AP IoU=0.50 with a value of 0.986, and AP IoU=0.75 with a value of 0.969. The trained SSD MobileNetV2 algorithm was then integrated into the developed Android application. The testing results on mid to high-end Android devices obtained an average inference time ranging from 165–230 ms. In addition, this application successfully detected waste objects according to those trained in the model. This application features a real-time scanning function with a classification mechanism for detected waste object types using bounding box colors, where organic waste is marked in green, inorganic waste in yellow, and B3 waste in red. With this mechanism, it provides an interactive experience for users in sorting their waste, thereby expected to increase awareness of waste sorting.]]>
Copyrights © 2025
