Back to search

Cornell University

Adding Model Uncertainty to Depth Prediction

Abstract

dc:description.abstract

Disparity and depth estimation of images is a fundamental problem for computer vision. Recent work has shown that convolutional neural networks are effective at both monocular and binocular depth prediction. However, standard neural networks do not give any information about the confidence of their predictions, making it impossible to know if a measurement could be inaccurate. In this work, we add Bayesian uncertainty to pretrained convolutional neural networks. Testing the networks on a synthetic dataset shows that the uncertainty is able to give confidence levels that are linked with the accuracy of the model output. Additionally, masking high uncertainty areas increases the remaining accuracy at the cost of decreasing the completeness of the output.

Degree

thesis:*
Name thesis:degree_name
M.S., Computer Science
Level thesis:degree_level
Master of Science
Discipline thesis:degree_discipline
Computer Science
Grantor
Cornell University
Year dc:date.issued
2019

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Wu, Eric
Chair dc:contributor.chair
  • Bala, Kavita
Committee member dc:contributor.committeemember
  • Easley, David Alan

Subjects

dc:subject × 1

Rights

Language dc:language.iso
en_US

Identifiers

dc:identifier.*
Dc Identifier Bibid
11050735
Dc Identifier Other
ProQuest Submission ID: 10659
ProQuest Publication ID: 22589613

Chain of custody

source
Harvested from
Cornell University
Base URL
ecommons.cornell.edu/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Wu, Eric. Adding Model Uncertainty to Depth Prediction. Master of Science thesis, Cornell University, 2019. https://hdl.handle.net/1813/67749