<![CDATA[Efficiency in waste management is a major challenge in modern cities. With so many people throwing away organic and inorganic waste, a solution is needed so that the waste can be sorted properly. Therefore, in this research, researchers aim to utilize computer vision technology based on the YOLOv9 algorithm to detect and sort organic and inorganic waste. Using a dataset of 6,747 images from the Roboflow platform, this system was trained to recognize various types of waste using the bounding box labeling method. The YOLOv9 algorithm is equipped with Programmable Gradient Information (PGI) and Generalized Efficient Layer Aggregation Network (GELAN) features, which provide superior performance in terms of accuracy and speed of system performance. The model training results show that YOLOv9 has a precision value of 0.83%, recall of 0.85%, and mAP of 0.8%, making the model reliable in detecting objects. However, there are several weaknesses, such as decreasing accuracy in blurry images, overlapping objects, and colors that have similar similarities, which can affect detection results by up to 20-30%. Compared to SSD MobileNet v2, YOLOv9 is superior in accuracy, precision and F-1 Score with results in Accuracy values of 58%, Precision 81%, F1-Score 69%. The Intersection over Union (IoU) test results produce excellent accuracy of 0.96%. This research recommends improvements through data augmentation and sensor integration to improve performance in various lighting conditions. This algorithm has great potential to be applied in technology-based waste management, supporting recycling efficiency, reducing human error, and providing a positive impact on the environment globally.]]>
Copyrights © 2025
