Tactile sensing has become increasingly important for robots, while most of the current tactile sensors— which measure force or pressure distribution— offer limited information. In this talk, I will introduce a novel tactile sensor, GelSight, which provide high spatial resolution images of the geometry of the contact surface. The sensor has a piece of painted elastomer on the surface, and an embedded camera to capture the shading change of the elastomer surface, which is used to reconstruct the 3D geometry of the contact surface. We also painted some markers on the sensor to infer general contact force and shear information. I will also introduce our research on using the sensor with robots on object recognition and estimating objects’ physical properties. The high spatial resolution of the sensor enables it to capture the very fine details on material surface, therefore to differentiate different material classes, or even different textiles for clothes. For physical property measurement, we showed that the sensor can estimate object hardness by pressing on it under a loosely controlled condition.