InsightGAN: Semi-Supervised Feature Learning with Generative Adversarial Network for Drug Abuse Detection

We present a novel generative adversarial network (GAN) model, called InsightGAN, for drug abuse detection. Our model is inspired by two closely related works on machine learning for healthcare applications: (1) drug abuse detection has been solved by machine learning with plentiful data from social media (where face pictures can be easily obtained); (2) facial characteristics have been explored in mental disorder diagnosis (drug addiction is also a mental disorder). In this paper, we adopt deep learning to extract discriminative facial features for drug abuse detection. However, in this application, the face pictures with ground-truth labels are far from sufficient for training a deep learning model. To alleviate the scarcity of labelled data, we thus propose a semi-supervised facial feature learning model based on GAN. Moreover, we also develop a robust algorithm for training our InsightGAN. Experimental results show the promising performance of our InsightGAN.