Talk map

(In progress) An Interactive Method for Generating Haptic Textures from Human Preferences
Mianlun Zheng, Shihan Lu, Matthew Fontaine (order not decided yet)

Abstract: A strong area of interest in the field of haptics is generating realistic textures. Many data-driven methods exist that can synthesize realistic feeling textures from recorded data. However, some soft and squishy textures (e.g. carpet or sponge) are difficult to record directly. Our work presents a first attempt of an interactive authoring system designed to generate textures purely from human preferences. In this interactive texture modeling system, Compositional Pattern Producing Network (CPPN) and Covariance Matrix Adaptation MAP-Elites (CMAME) are used for texture modeling and interactive space searching, respectively. Results from a two-part user study show our method has a positive effect on generation and recognition of rough and coarse textures, though it fails to generate some easily recognizable textures. These results suggest that this line of research is promising and our report details the current limitations of our system with a road-map for addressing those limitations in future work.