Poster: PressureSilhouette: A Model to Predict Sleep Silhouette from Pressure Mattress Images

Sleep monitoring has many health applications. One of the main tasks in sleep monitoring is sleep posture classification. Most of the traditional approaches have a camera as their primary detector which can be privacy-invasive and malfunction in some settings. An alternative is to use pressure mats. The problem however with most of the pressure mat raw images is that they contain gaps and discontinuous blocks which makes direct classification challenging. To address this issue, we propose PressureSilhouette, a deep learning model based on the Pix2Pix model that predicts the silhouette of the body. Using the generated silhouette, we will be able to predict the posture classification more accurately.