DeepMVS: Learning Multi-view Stereopsis –Supplementary Material–

In this supplementary document, we present additional implementation details and results to complement the main paper. We first describe in detail the formulation for applying DenseCRF to disparity refinement. We then present additional experimental results, including evaluation using scale-invariant errors, results on DeMoN’s testing dataset, and the effect of input image resolutions. Finally, we show additional qualitative comparisons with conventional MVS algorithms.