Multi-View Stereo by Temporal Nonparametric Fusion

Yuxin Hou, Juho Kannala, Arno Solin; Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2019, pp. 2651-2660


We propose a novel idea for depth estimation from multi-view image-pose pairs, where the model has capability to leverage information from previous latent-space encodings of the scene. This model uses pairs of images and poses, which are passed through an encoder-decoder model for disparity estimation. The novelty lies in soft-constraining the bottleneck layer by a nonparametric Gaussian process prior. We propose a pose-kernel structure that encourages similar poses to have resembling latent spaces. The flexibility of the Gaussian process (GP) prior provides adapting memory for fusing information from nearby views. We train the encoder-decoder and the GP hyperparameters jointly end-to-end. In addition to a batch method, we derive a lightweight estimation scheme that circumvents standard pitfalls in scaling Gaussian process inference, and demonstrate how our scheme can run in real-time on smart devices.

Related Material

[pdf] [supp]
author = {Hou, Yuxin and Kannala, Juho and Solin, Arno},
title = {Multi-View Stereo by Temporal Nonparametric Fusion},
booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
month = {October},
year = {2019}