Convolutional Neural Network for Identification of Personal Protective Equipment Usage Compliance in Manufacturing Laboratory

Data from the Badan Penyelenggara Jaminan Sosial (BPJS) Ketenagakerjaan Indonesia 2019 to 2021 shows that number of work accident victims who claimed Work Accident Insurance (Jaminan Kecelakaan Kerja / JKK) continues increase. The high accidents is mostly caused by unsafe behavior at sites, one which in terms compliance with use Personal Protective Equipment (PPE). One tool considered important as a step reducing an identification system for personal protective safety equipment can detect PPE used visitors or workers. This study develops automatic built using Convolutional Neural Network (CNN) identify manufacturing technology laboratory. CNN models are 4th and 5th versions You Only Look Once (YOLO) then compared based on two methods: train scratch transfer learning. dataset building detection has 11,579 images consisting six classes objects. Overall performance proposed very good results. Moreover, comparison result among three YOLOv5 learning best precision (94.2 %), recall (91.8 mAP (88.6%).