Superb Monocular Depth Estimation Based on Transfer Learning and Surface Normal Guidance
| Autor: | Qu Xingtian, Zhang Wang, Zhen Chen, Qi Haogang, Zhao Fengshang, Huang Kang, Shouqian Chen |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2020 |
| Předmět: |
Computer science
Point cloud ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Multi-task learning multi-task learning monocular depth estimation 02 engineering and technology Simultaneous localization and mapping transfer learning lcsh:Chemical technology 01 natural sciences Biochemistry Convolutional neural network Article Analytical Chemistry SFM 0202 electrical engineering electronic engineering information engineering Computer vision lcsh:TP1-1185 Motion planning Electrical and Electronic Engineering surface normal estimation Instrumentation Monocular business.industry Deep learning supervised deep learning 010401 analytical chemistry Atomic and Molecular Physics and Optics 0104 chemical sciences SLAM 020201 artificial intelligence & image processing Artificial intelligence business Normal |
| Zdroj: | Sensors, Vol 20, Iss 4856, p 4856 (2020) Sensors Volume 20 Issue 17 Sensors (Basel, Switzerland) |
| ISSN: | 1424-8220 |
| Popis: | Accurately sensing the surrounding 3D scene is indispensable for drones or robots to execute path planning and navigation. In this paper, a novel monocular depth estimation method was proposed that primarily utilizes a lighter-weight Convolutional Neural Network (CNN) structure for coarse depth prediction and then refines the coarse depth images by combining surface normal guidance. Specifically, the coarse depth prediction network is designed as pre-trained encoder&ndash decoder architecture for describing the 3D structure. When it comes to surface normal estimation, the deep learning network was designed as a two-stream encoder&ndash decoder structure, which hierarchically merges red-green-blue-depth (RGB-D) images for capturing more accurate geometric boundaries. Relying on fewer network parameters and simpler learning structure, better detailed depth maps are produced than the existing states. Moreover, 3D point cloud maps reconstructed from depth prediction images confirm that our framework can be conveniently adopted as components of a monocular simultaneous localization and mapping (SLAM) paradigm. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|