Perbandingan Performa Model SSD Mobilenet V2 dan FPNLite dalam Deteksi Helm Pengendara Sepeda Motor

Abstract

One important aspect in computer vision is object detection, which aims to identify and determine the position of objects in images. In the context of safety, detecting helmet-wearing objects in motorcycle riders is crucial to reduce the risk of accidents and protect the riders. Helmets are the primary protective gear for motorcycle riders, safeguarding their heads from serious injuries during accidents. In this research, we implemented helmet object detection using the TensorFlow Framework with pre-trained models based on the Single Shot Multibox Detector (SSD) architecture, specifically the Mobilenet V2 and Mobilenet V2 FPNLite models. The Mobilenet V2 and Mobilenet V2 FPNLite models were trained using a dataset consisting of images of motorcycle riders wearing helmets and not wearing helmets. The performance evaluation results of both models using the mean Average Precision (mAP) metric showed that the proposed model achieved an mAP of 71.59% for the Mobilenet V2 FPNLite model and 80.12% for the Mobilenet V2 model.

Keywords – Object Detection, Helmet, Tensorflow, SSD, Imagery