Implementation of Tableware Detection Based on YOLO V4

The ability of object detection is one of the largest enablers which will give the robot the input to make some autonomous decisions and is particularly important for robots assisting humans. YOLO (You Only Look Once), as an object detection system, is known as its optimal accuracy and real time. The purpose of this paper was to realize the object detection of tableware by using the YOLO algorithm. In this paper, firstly, 584 images which include plates, bowls, cups, chopsticks, spoons, forks as well as knives used for training and testing were collected from the Internet; secondly these images were labelled by the makesense.ai; finally, these images were divided into training dataset used for training the YOLO V4 model and testing dataset used for verifying the validation of the model. The training and testing results indicated that YOLO V4 can achieve high accuracy of tableware detection.