抄録
The goal of this work was to create a system to find the medical errors in prescribing One-Dose Package (ODP) systems by teaching the template package of ODP as correct and comparing the content of other ODPs. This study is a preliminary step in the construction of such a system, in which each drug is recognized separately. The first step of this work is to use YOLOv5 to detect the color of pills. Second, features were extracted from each pill. A database of the pills and features was created. In the last step, a support vector machine (SVM) is run to classify the pills. The result-of-colored pill was classified perfectly using the created database, but the accuracy of the classification for white pills reached only 66%. The best features to classify the colored pills were the a* and b* values of the L*a*b* color space.
https://ror.org/03hv1ad10 
