Detection and Identification of Pills using Machine Learning Models

Abstract

Pill color, pill size and shape are few important characteristics for automatic pill detection. However, the environmental factors may cause an effect such that a change is produced in the above factors. Often medication errors occur that may cause complications to patients and all these are caused due to damage in labels and mismatches in medicine intake, etc. In this report, a trained system is proposed using Keras and Tensor Flow mainly, for easy and quick identification of varieties of pills. The detected pill (object detection) connects to the pill database wherein the pill name is detected. After the process of detection, the pre-trained dataset is used to identify the pill. Moreover, the dataset would also include the use cases and required detailed information of the respective pill. The project involves collecting datasets for automated medicine detecting technology. Effectiveness of the proposed method can be verified in the experimental results.