Deep learning is a strong tool for predicting scene properties from images. Typical supervised methods require large scale real data with ground truth, which is hard to obtain. This situation demands techniques using little ground truth real data. Without annotations, an apparent question is: Where does the supervision signal come from for training deep networks? In this talk, we demonstrate that the awareness of materials provides such easy-to-obtain signals. We propose a framework that can be used for different tasks to exploit material-aware supervisions. To demonstrate the effectiveness of the proposed framework, we present four applications, including translucent powder recognition, human geometry and texture reconstruction, floor appearance decomposition for object insertion, and cross-spectral stereo matching while fixing non-Lambertian regions.
Srinivasa G. Narasimhan (Co-Chair)
Martial Hebert (Co-Chair)
Matthew P. O’Toole
Sing Bing Kang (Zillow Group)
Please remember to adhere to university in person policy.
Zoom Participation. See announcement.