Applying Google's Teachable Machine to Detect the Faces of Criminals according to Arrest Warrants

Abstract

This research focuses on studying and evaluating the feasibility of using artificial intelligence (AI), specifically Google’s Teachable Machine, for detecting faces of criminals in the border areas of three southern provinces. The researchers conducted a study and research process involving the examination of three hyperparameters of Google’s Teachable Machine: 1) Epochs, 2) Batch size, and 3) Learning rate. Data preparation and system testing were carried out by inputting facial data of terrorist and non- terrorist to allow the AI to learn and differentiate between the two, a process known as machine learning. The researchers sought appropriate hyperparameter values through study and iterative experimentation to achieve the clearest and least biased face detection possible. The research findings revealed that the suitable hyperparameter values were Epochs = 100, Batch size = 256, and Learning rate = 0.001. Subsequent testing demonstrated that Google’s Teachable Machine achieved a face detection accuracy of 96.50%. This study illustrates that Google’s Teachable Machine is effective in face detection and serves as a valuable tool, saving time and reducing programming complexities. Individuals interested in further exploration can utilize Google’s Teachable Machine or AI to obtain more accurate and efficient results.