We present a new multi-sensor dataset for multi-view 3D surface reconstruction. It includes registered RGB and depth data from sensors of different resolutions and modalities: smartphones, Intel RealSense, Microsoft Kinect, industrial cameras, and structured-light scanner. The data for each scene is obtained under a large number of lighting conditions, and the scenes are selected to emphasize a diverse set of material properties challenging for existing algorithms. Overall, we provide around 1.4 million images of 107 different scenes acquired at 14 lighting conditions from 100 viewing directions. We expect our dataset will be useful for evaluation and training of 3D reconstruction algorithms of different types and for other related tasks.