Depth estimation of a single RGB image with semisupervised two-stage regression

Jun Chi, Jie Gao, Lin Qi, Shu Zhang, Junyu Dong, Hui Yu

Research output: Chapter in Book/Report/Conference proceedingConference contribution

61 Downloads (Pure)

Abstract

Obtaining accurate depth estimation at low computational cost is a major problem in the field of computer vision. To tackle this problem, we propose a framework that integrates different neural networks, for predicting the corresponding depth from a single RGB image and sparse depth samples. This method combines two different types of deep learning frameworks with the best performance, including the improved Residual Neural Network and conditional generation adversarial network (cGAN). It has been proved that the improved ResNet has strong depth prediction capability, but the depth map is still incomplete in detail. We improve the existing cGAN model to enhance ResNet-based depth prediction. Experiments compared with stage-of-the-art are performed on publicly available data sets. And the results demonstrate that the proposed two-stage deep regression model is superior to other existing methods of the same type.
Original languageEnglish
Title of host publicationICCIP '19: Proceedings of the 5th International Conference on Communication and Information Processing
PublisherAssociation for Computing Machinery
Number of pages5
ISBN (Print)978-1-4503-7258-9
DOIs
Publication statusPublished - 15 Nov 2019
Event2019 the 5th International Conference on Communication and Information Processing - Chongqing, China
Duration: 15 Nov 201917 Nov 2019

Conference

Conference2019 the 5th International Conference on Communication and Information Processing
Abbreviated titleICCIP 2019
CountryChina
CityChongqing
Period15/11/1917/11/19

Keywords

  • CCS
  • computing methodologies
  • coputer graphics
  • image manipulation
  • image processing

Fingerprint

Dive into the research topics of 'Depth estimation of a single RGB image with semisupervised two-stage regression'. Together they form a unique fingerprint.

Cite this