Recognition of Tooth Types and Tooth Conditions Using YOLO and Combination WLD&HOG - SVM

This paper presents the development of a dentistry medical record call as Odontogram. The majority of odontogram medical record data is still entered manually on paper. It takes too long to finish just one evaluation when a dentist fills out documents manually by drawing and identifying the state of each patient's tooth based on his or her observations. We build a system to identify tooth types using YOLOv5 and recognize tooth conditions using a combination of WLD & HOG classified with SVM to help doctors analyze patient tooth. This paper classifies eight types of teeth: incisors, incisors, canines, premolars 1, premolars 2, molars 1, molars 2 and molars 3 as well as five types of dental conditions: normal, filling, caries, radix gangrene, and impaction. Based on our experiments, YOLOv5 achieves an accuracy of 93 % and the combination of WLD & HOG and SVM achieves an accuracy of $77,6 \%$.