Sumeth YuenyongNarit HnoohomKonlakorn WongpatikasereeMahidol University2019-08-232019-08-232018-06-081st International ECTI Northern Section Conference on Electrical, Electronics, Computer and Telecommunications Engineering, ECTI-NCON 2018. (2018), 65-682-s2.0-85049945756https://repository.li.mahidol.ac.th/handle/20.500.14594/45626© 2018 IEEE. The researchers present an experiment on automatic detection of concealed knives in infrared (IR) images. The researchers generated a dataset, called the IR dataset, which contained 8,527 images. The dataset was divided into two groups of IR images comprised of person without knife and person with a knife. Knives of different shapes and sizes were concealed under normal clothing, then images of person subjects carrying hidden knives were taken with a smartphone IR camera add-on. A deep neural network that was trained on natural image (GoogleNet dataset) was fine-tuned to classify the IR images as person, or person carrying hidden knife. The classification accuracy on a separate test set shows that hidden knives can be detected at 97.91% accuracy.Mahidol UniversityComputer ScienceEngineeringConference PaperSCOPUS10.1109/ECTI-NCON.2018.8378283