Keselamatan dan kesehatan kerja (K3) menjadi prioritas utama di industri minyak dan gas, termasuk PT Pertamina EP – Poleng Field yang memiliki aktivitas bongkar muat berisiko tinggi. Pengawasan manual melalui CCTV dan dokumen perizinan kerja belum mampu memberikan deteksi real-time terhadap keberadaan pekerja dan kendaraan berat, sehingga berpotensi meningkatkan risiko kecelakaan kerja. Penelitian ini bertujuan mengimplementasikan algoritma You Only Look Once versi 8s (YOLOv8s) untuk mendeteksi objek person dan truck secara otomatis pada lingkungan kerja. Metode penelitian meliputi pengumpulan dataset, anotasi menggunakan Oriented Bounding Box (OBB), pelatihan model YOLOv8s-OBB dengan konfigurasi hyperparameter tertentu, dan evaluasi kinerja menggunakan metrik precision, recall, F1-score, dan mean Average Precision (mAP). Hasil pengujian menunjukkan kinerja tinggi dengan precision 0,999, recall 0,992, mAP50 sebesar 0,993, dan mAP50–95 sebesar 0,934 secara keseluruhan. Pada kelas person, diperoleh precision 1 dan recall 0,984, sedangkan pada kelas truck, precision mencapai 0,998 dan recall 1. Implementasi ini diharapkan dapat meningkatkan efisiensi pengawasan, mengurangi risiko kecelakaan kerja, serta mendukung penerapan ISO 45001 dan tujuan Sustainable Development Goals (SDG) 9 di sektor industri minyak dan gas.

Occupational health and safety (OHS) is a top priority in the oil and gas industry, including at PT Pertamina EP – Poleng Field, where high-risk loading and unloading activities are common. Manual monitoring through CCTV and work permit documentation has not yet provided real-time detection of workers and heavy vehicles, potentially increasing the risk of workplace accidents. This study aims to implement the You Only Look Once version 8s (YOLOv8s) algorithm to automatically detect person and truck objects in the work environment. The research method includes dataset collection, annotation using the Oriented Bounding Box (OBB) format, training the YOLOv8s-OBB model with specific hyperparameter configurations, and evaluating its performance using precision, recall, F1-score, and mean Average Precision (mAP) metrics. Testing results showed high performance with an overall precision of 0.999, recall of 0.992, mAP50 of 0.993, and mAP50–95 of 0.934. For the person class, precision was 1 and recall 0.984, while for the truck class, precision reached 0.998 and recall 1. This implementation is expected to enhance monitoring efficiency, reduce workplace accident risks, and support ISO 45001 implementation as well as Sustainable Development Goals (SDG) 9 in the oil and gas sector.