Object Detection using NAO Humanoid Robot Based on YOLO Model

Abstract

Object detection is an important part in the field of robotics as it enables robots to understand their surroundings. The NAO humanoid robot is extensively used in human-robot interaction research. In this study a novel approach combining the VGG16 network and You Only Look Once (YOLO) algorithm is used for object detection using the NAO robot. YOLOv7 is selected for its best balanced information retention, quick inference, accurate localization, and identification of objects as compared to several bounding box algorithms. VGG16 network is adopted as a feature extractor to optimize the performance of object detection for NAO low-resolution camera images. Once feature extraction is completed then it’s output layer is combined with our fine tuned YOLOv7 model for object detection. The fine-tuned YOLOv7 model is proposed with some pre-processing techniques such as image augmentation, angle movement, and scale resizing for the performance improvement. The efficiency of the proposed model is compared with the performance of other state-of-the-art models.