Decoupled and Reparameterized Compound Attention-Based Light Field Depth Estimation Network
A light field (LF) camera captures both spatial and angular information of the real world, and the intertwined nature of these dimensions presents a pressing challenge in effectively disentangling meaningful LF information for depth estimation. This paper introduces a feature extraction network base...
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2023-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10322864/ |
| _version_ | 1797473491524517888 |
|---|---|
| author | Wan Liao Xiaoqi Bai Qian Zhang Jie Cao Haoyu Fu Wei Wei Bin Wang Tao Yan |
| author_facet | Wan Liao Xiaoqi Bai Qian Zhang Jie Cao Haoyu Fu Wei Wei Bin Wang Tao Yan |
| author_sort | Wan Liao |
| collection | DOAJ |
| description | A light field (LF) camera captures both spatial and angular information of the real world, and the intertwined nature of these dimensions presents a pressing challenge in effectively disentangling meaningful LF information for depth estimation. This paper introduces a feature extraction network based on LF decoupling, which ingeniously separates the LF. Furthermore, given the extensive volume of input data inherent in LF images, a novel reparameterizable Residual-Densely Branched Leaky-ReLU Block(Res-DBLB) architecture was developed to replace conventional residual structures and multibranch architectures and enhance inference efficiency. Incorporating an attention mechanism further refines the network, effectively addressing the computational intensity and time-consuming nature of LF depth estimation, thus furthering the advancement of this technology. Our model was applied to widely used datasets as well as the latest LF datasets, namely HCI and UrbanLF, showing superior performance over six other popular models across four evaluation metrics. |
| first_indexed | 2024-03-09T20:15:14Z |
| format | Article |
| id | doaj.art-254ba6fe9a844946a0664f63d9f44c8b |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-03-09T20:15:14Z |
| publishDate | 2023-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-254ba6fe9a844946a0664f63d9f44c8b2023-11-24T00:00:54ZengIEEEIEEE Access2169-35362023-01-011113011913013010.1109/ACCESS.2023.333464010322864Decoupled and Reparameterized Compound Attention-Based Light Field Depth Estimation NetworkWan Liao0https://orcid.org/0000-0002-1300-3580Xiaoqi Bai1Qian Zhang2https://orcid.org/0000-0003-0760-9241Jie Cao3Haoyu Fu4Wei Wei5Bin Wang6Tao Yan7https://orcid.org/0000-0002-8304-8733College of Information, Mechanical and Electrical Engineering, Shanghai Normal University, Shanghai, ChinaEducation Institute of Yangpu District Shanghai, Shanghai, ChinaCollege of Information, Mechanical and Electrical Engineering, Shanghai Normal University, Shanghai, ChinaCollege of Information, Mechanical and Electrical Engineering, Shanghai Normal University, Shanghai, ChinaCollege of Information, Mechanical and Electrical Engineering, Shanghai Normal University, Shanghai, ChinaCollege of Information, Mechanical and Electrical Engineering, Shanghai Normal University, Shanghai, ChinaCollege of Information, Mechanical and Electrical Engineering, Shanghai Normal University, Shanghai, ChinaSchool of Mechanical, Electrical and Information Engineering, Putian University, Putian, ChinaA light field (LF) camera captures both spatial and angular information of the real world, and the intertwined nature of these dimensions presents a pressing challenge in effectively disentangling meaningful LF information for depth estimation. This paper introduces a feature extraction network based on LF decoupling, which ingeniously separates the LF. Furthermore, given the extensive volume of input data inherent in LF images, a novel reparameterizable Residual-Densely Branched Leaky-ReLU Block(Res-DBLB) architecture was developed to replace conventional residual structures and multibranch architectures and enhance inference efficiency. Incorporating an attention mechanism further refines the network, effectively addressing the computational intensity and time-consuming nature of LF depth estimation, thus furthering the advancement of this technology. Our model was applied to widely used datasets as well as the latest LF datasets, namely HCI and UrbanLF, showing superior performance over six other popular models across four evaluation metrics.https://ieeexplore.ieee.org/document/10322864/Light field depth estimationdecouplingreparameterizationattention mechanism |
| spellingShingle | Wan Liao Xiaoqi Bai Qian Zhang Jie Cao Haoyu Fu Wei Wei Bin Wang Tao Yan Decoupled and Reparameterized Compound Attention-Based Light Field Depth Estimation Network IEEE Access Light field depth estimation decoupling reparameterization attention mechanism |
| title | Decoupled and Reparameterized Compound Attention-Based Light Field Depth Estimation Network |
| title_full | Decoupled and Reparameterized Compound Attention-Based Light Field Depth Estimation Network |
| title_fullStr | Decoupled and Reparameterized Compound Attention-Based Light Field Depth Estimation Network |
| title_full_unstemmed | Decoupled and Reparameterized Compound Attention-Based Light Field Depth Estimation Network |
| title_short | Decoupled and Reparameterized Compound Attention-Based Light Field Depth Estimation Network |
| title_sort | decoupled and reparameterized compound attention based light field depth estimation network |
| topic | Light field depth estimation decoupling reparameterization attention mechanism |
| url | https://ieeexplore.ieee.org/document/10322864/ |
| work_keys_str_mv | AT wanliao decoupledandreparameterizedcompoundattentionbasedlightfielddepthestimationnetwork AT xiaoqibai decoupledandreparameterizedcompoundattentionbasedlightfielddepthestimationnetwork AT qianzhang decoupledandreparameterizedcompoundattentionbasedlightfielddepthestimationnetwork AT jiecao decoupledandreparameterizedcompoundattentionbasedlightfielddepthestimationnetwork AT haoyufu decoupledandreparameterizedcompoundattentionbasedlightfielddepthestimationnetwork AT weiwei decoupledandreparameterizedcompoundattentionbasedlightfielddepthestimationnetwork AT binwang decoupledandreparameterizedcompoundattentionbasedlightfielddepthestimationnetwork AT taoyan decoupledandreparameterizedcompoundattentionbasedlightfielddepthestimationnetwork |