Arwin Datumaya Wahyudi Sumari, Ndaru Atmi Purnami, Bangga Dirgantara, Iqmal Agil Alfian, Ilham Yoga Pratama, Dian Permana, Denny Dermawan, Ika Noer Syamsiana
Datasets are a critical component of an Artificial Intelligence-based system, specifically Machine Learning and its derivatives, such as deep learning. Datasets are the primary requirement to build knowledge about a phenomenon faced in the real world. This research is part of designing and implementing the Human Following Robot (HFR), a robot designed to follow humans with specific tasks assigned to them. Our designed and implemented HFR is intended to help aircraft technicians or mechanics perform aircraft maintenance in shelters that can be very far from the storage room of aircraft maintenance equipment by bringing the aircraft maintenance equipment or tool kit. We have developed the DINI dataset to enable the HFR to assist the Indonesian Air Force technicians effectively. This custom dataset stores image data from the typical uniform color of the Indonesian Air Force personnel, namely the blue color, which was built specifically for HFR. The dataset development involved finding the image data sources, data collection, dataset creation using Roboflow tool, and dataset testing. The created dataset with 1,451 images is vital to train and test HFR's ability to recognize blue uniforms. To measure its performance, the dataset was tested using YOLO version 5n (YOLOv5n). The results show that the DINI dataset achieved a Precision value of 100%, a Recall value of 42%, an Accuracy value of 87%, and an F1-score value of 59%. These results also show that the DINI dataset can be enhanced to accommodate other colors as needed in the field. © 2025 IEEE.
State Polytechnic of Malang, Department of Electrical Engineering, Malang, Indonesia; Adisutjipto Institute of Aerospace Technology, Faculty of Industrial Technology, Yogyakarta, Indonesia; Adisutjipto Institute of Aerospace Technology, Faculty of Aerospace Technology, Yogyakarta, Indonesia