Real-Time Pill Identification with Prescription Confirmation Using Deep Learning on Embedded System

In the study, a real-time pill identification with prescription confirmation within medical packaging using deep learning on embedded system is proposed. In this system, multiple pills within a medical packaging for inpatients can be identified using deep learning and confirmed with the doctor’s prescription in QR code to avoid error in the packaging process. The system is implemented in an embedded system with GPU for accelerating the identification and confirmation processes. Experimental results of the proposed system using Nvidia’s Jetson Nano shown that multiple pills with different color and shapes can be identified and confirmed with doctor’s prescription in real time.