A Dataset of Thermal Interfaces

Interfaces based on thermal imaging coined as thermal interfaces could be used to create unique, individual experiences by allowing users to create personalized actions on their preferred interfaces that are not tied to a specific location. For this, either thermal cameras could be installed in the user's environment or thermal cameras could be integrated into wearables, such as head-mounted displays (HMDs). Previous work already investigated different surfaces for thermal interfaces based on very basic interactions and complex interactions remain unexplored. To offer interfaces based on thermal imaging, we need simple ways for users to enter information as well as an option to automatically recognize the information entered by the users. To do so, we collected thermal images of gesture data from 32 participants on four different surfaces which are plastic, wood, paper, and a mirror. Further, we investigated three different types of gestures which are simple lines without overlapping, lines with one overlapping, and lines that overlap twice. Each gesture was recorded by two thermal cameras (FLIR E8-XT and Optris PI 450i).

The dataset consists of 384 thermal images with 12 different gesture type recorded using the FLIR camera and another 384 thermal images from the Optris camera. The ground truth gesture type and details about surfaces are included in the attached excel sheet.