Matsushita Lab.

Deep learning-based photometric stereo

This work is the first photometric stereo method based on deep learning. Our method uses a deep neural network to model reflectances in the real world. As a result, our method achieved the best results compared to conventional methods in DiLiGenT Benchmark comparison. [Project Page]


  • Hiroaki Santo, Masaki Samejima, Yusuke Sugano, Boxin Shi, and Yasuyuki Matsushita: Deep photometric stereo network, International Workshop on Physics Based Vision meets Deep Learning (PBDL) in Conjunction with IEEE International Conference on Computer Vision (ICCV), Venice, Italy (Oct. 2017).
Matsushita Lab.