Computergrafik

Two-shot Spatially-varying BRDF and Shape Estimation

Mark Boss1, Varun Jampani2, Kihwan Kim2, Hendrik P. A. Lensch1, and Jan Kautz2
University of Tübingen1 , NVIDIA2
Conference on Computer Vision and Pattern Recognition (CVPR) 2020

Abstract

Capturing the shape and spatially-varying appearance (SVBRDF) of an object from images is a challenging task that has applications in both computer vision and graphics. Traditional optimization-based approaches often need a large number of images taken from multiple views in a controlled environment. Newer deep learning-based approaches require only a few input images, but the reconstruction quality is not on par with optimization techniques. We propose a novel deep learning architecture with a stage-wise estimation of shape and SVBRDF. The previous predictions guide each estimation, and a joint refinement network later refines both SVBRDF and shape. We follow a practical mobile image capture setting and use unaligned two-shot flash and no-flash images as input. Both our two-shot image capture and network inference can run on mobile hardware. We also create a large-scale synthetic training dataset with domain-randomized geometry and realistic materials. Extensive experiments on both synthetic and real-world datasets show that our network trained on a synthetic dataset can generalize well to real-world images. Comparisons with recent approaches demonstrate the superior performance of the proposed approach.

Links

Bibtex

@inproceedings{Boss2020-TwoShotShapeAndBrdf,
    author = {Boss, Mark and Jampani, Varun and Kim, Kihwan and Lensch, Hendrik P.A. and Kautz, Jan},
    title = {Two-shot Spatially-varying BRDF and Shape Estimation},
    booktitle = {IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
    year = {2020}
}

Copyright

This material is posted here with permission of the IEEE. Internal or personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution must be obtained from the IEEE by writing to pubs-permissionsspam prevention@ieee.org.