PharmaDetect: Drug Identification and Guidance through Visual Recognition

This study proposes an approach to drug identification using image recognition and machine learning. Aiming to address medication misuse arising from unclear prescriptions and faded labels, this study develops a program that recognizes drugs based on their external features like size, shape, and color. A classification model is employed to train on features extracted from images of 10 different medicine types. Through 5-fold cross-validation, the Support Vector Machine achieved an average test fold accuracy of 94%, with a 100% accuracy rate for identifying new medicine images taken under usual condition. This research demonstrates the feasibility of utilizing image recognition and machine learning for accurate medication identification and knowledge retrieval, potentially contributing to improved medication adherence and reduced instances of misuse.