Skip to main navigation Skip to search Skip to main content

Random Mask Slice Stitching(RMSS): A Data Augmentation for Depth Estimation

  • Beihang University
  • Zhejiang Mobile Information System Integration Co.,Ltd

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

This work aims to summarize some existing data augmentation methods in the field of depth estimation, and propose a depth estimation data augmentation method suitable for both supervised learning and self-supervised learning. In the training task of computer vision, data augmentation is very significant to improve the accuracy and robustness of the model. However, for depth estimation tasks, data augmentation is not as common as detection or segmentation tasks. Although some data enhancement methods have been proposed in recent years, these data enhancement methods are not effective in different datasets or in supervised learning modes and unsupervised learning modes. Therefore, the data augmentation methods proposed in recent years are sorted out in this paper, mainly for two challenging datasets NYU-Depth-v2 and KITTI. In this paper we propose a simple and effective data augmentation method Random Mask Slice Stitching(RMSS) that can be used in supervised and self-supervised tasks and outperforms existing methods.

Original languageEnglish
Title of host publicationProceedings of the 35th Chinese Control and Decision Conference, CCDC 2023
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages870-875
Number of pages6
ISBN (Electronic)9798350334722
DOIs
StatePublished - 2023
Event35th Chinese Control and Decision Conference, CCDC 2023 - Yichang, China
Duration: 20 May 202322 May 2023

Publication series

NameProceedings of the 35th Chinese Control and Decision Conference, CCDC 2023

Conference

Conference35th Chinese Control and Decision Conference, CCDC 2023
Country/TerritoryChina
CityYichang
Period20/05/2322/05/23

Keywords

  • Computer vision
  • Convolutional neural network
  • Data augmentation
  • Depth estimation
  • Transformer

Fingerprint

Dive into the research topics of 'Random Mask Slice Stitching(RMSS): A Data Augmentation for Depth Estimation'. Together they form a unique fingerprint.

Cite this