Implementasi Deteksi Objek untuk Pemantauan Kesehatan Gigi berbasis Sistem Tertanam
SAYYIDAN MUHAMAD IKHSAN, Dr. Eng. Ir. Igi Ardiyanto, S.T., M.Eng., SMIEEE; Ir. Prapto Nugroho, S.T., M.Eng., D.Eng., IPM.
2024 | Skripsi | S1 TEKNIK BIOMEDIS
Dental health problems are a significant issue in Indonesia, with dental caries being a major concern, especially among young children, where the prevalence reaches 93%. Although regular dental check-ups and treatments can address most dental health issues, limited access and high costs of dental services are major barriers, particularly in remote, underdeveloped, and outermost regions (3T areas). Studies indicate that artificial intelligence (AI) can be used for self-detection of dental health, reducing costs and achieving performance comparable to or better than dental specialists. However, AI requires intensive computation, making its implementation and development expensive and challenging, thus less practical for widespread use. Therefore, this research examines the application of object detection technology on embedded systems for dental health monitoring. The aim is to implement a dental health detection system on embedded systems, with the hope of providing a reference for the development of economical and efficient intelligent detection systems in the future.
The methodology used in this research includes several stages, namely problem definition, solution design, data collection and configuration, model training, model conversion, model validation, and model implementation on the embedded system. This research uses two algorithms and two different resolutions, namely FastestDet and YOLOv8n with resolutions of 352x352 pixels and 160x160 pixels. The results show that object detection can be implemented well on embedded systems by using lightweight object detection algorithms. The 352x352 pixel resolution and YOLOv8 algorithm are superior as they offer higher detection performance with insignificant difference in inference time and FPS.
This research demonstrates that object detection can be effectively implemented on embedded systems with limited computational resources, with a trade-off to be considered between inference speed and detection accuracy.
Kata Kunci : FastestDet, Sistem Tertanam, Deteksi Objek, YOLOv8, Gigi