<![CDATA[The increasing presence of inorganic waste such as plastic bottles and cans poses a serious environmental threat and demands efficient detection solutions. This study aims to develop a real-time detection and position estimation system for inorganic waste using the YOLO v4-Tiny algorithm combined with homography transformation. A total of 3,388 labeled images were prepared and augmented via the Roboflow platform. The detection model was trained using the Darknet framework, while homography was applied to estimate object positions in real-world coordinates. System performance was evaluated based on precision, recall, F1-score, mean Average Precision (mAP), and Intersection over Union (IoU). The results show a mAP of 86.43%, precision of 77%, recall of 90%, and an average IoU of 62.33%. The system achieved a frame rate of 3–5 FPS, demonstrating its potential for low-power embedded devices. This approach is suitable for real-time waste monitoring using computer vision in constrained environments.]]>
Copyrights © 2025
