DeepMVS: learning multi-view stereopsis

We present DeepMVS, a deep convolutional neural network (ConvNet) for multi-view stereo reconstruction. Taking an arbitrary number of posed images as input, we first produce a set of plane-sweep volumes and use the proposed DeepMVS network to predict high-quality disparity maps. The key contributions that enable these results are (1) supervised pretraining on a photorealistic synthetic dataset, (2) an effective method for aggregating information across a set of unordered images, and (3) integrating multi-layer feature activations from the pre-trained VGG-19 network. We validate the efficacy of DeepMVS using the ETH3D Benchmark. Our results show that DeepMVS compares favorably against state-of-the-art conventional MVS algorithms and other ConvNet based methods, particularly for near-textureless regions and thin structures.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Electrical & Computer Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2018

Author dc:creator

• Huang, Po-Han

Contributors dc:contributor

• Ahuja, Narendra

• computer vision, deep learning, multi-view stereo, DeepMVS, pattern recognition, 3D reconstruction, depth estimation, machine learning, convolutional neural network

Statement dc:rights

• Copyright 2018 Po-Han Huang

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/101770

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/101770