Article
Computer Science, Information Systems
Weiqiang Wang, Chao Tan, Yunbing Yan
Summary: Traditional CNN often fails to capture global context information effectively in the process of environmental perception due to its network structure, resulting in blurred edges of objects and scenes. To address this issue, a self-supervised monocular depth estimation algorithm incorporating a Transformer is proposed. Experimental results show that the proposed algorithm outperforms mainstream algorithms, reducing the absolute relative error by 3.7% and the squared relative error by 3.9% compared to the latest CNN-Transformer algorithm.
Article
Computer Science, Artificial Intelligence
Chunpu Liu, Wangmeng Zuo, Guanglei Yang, Wanlong Li, Feng Wen, Hongbo Zhang, Tianyi Zang
Summary: Monocular depth estimation plays a crucial role in 3D scene understanding and has gained increasing attention in the computer vision field. Deep learning-based monocular depth estimation methods have achieved significant performance by exploring various network architectures. However, leveraging scene geometry relations to improve the performance of these models has been less studied. In this paper, we propose a geometry-aware constraint that utilizes relative order information to enhance monocular depth estimation models.
APPLIED INTELLIGENCE
(2023)
Article
Engineering, Electrical & Electronic
Anmei Zhang, Yunchao Ma, Jiangyu Liu, Jian Sun
Summary: Deep learning approach has achieved great success in monocular depth estimation. However, the learned deep network may produce a depth map with fewer details and incorrect global depth layout, especially when the learned network is applied to a high-resolution image. Our proposed multi-scale residual Laplacian pyramid fusion net (MS-RLap-FNet) aims to generate a high-quality depth map by refining the depth maps estimated by existing models.
IEEE SIGNAL PROCESSING LETTERS
(2023)
Article
Computer Science, Artificial Intelligence
Xibin Song, Wei Li, Dingfu Zhou, Yuchao Dai, Jin Fang, Hongdong Li, Liangjun Zhang
Summary: This paper proposes a novel MLDA-Net framework for self-supervised depth estimation, improving the quality of depth maps through multi-level feature extraction and dual-attention strategy.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Computer Science, Hardware & Architecture
Yi Yang, Lihua Tian, Chen Li, Botong Zhang
Summary: In addition to RGB information, depth information is crucial in image analysis. This paper proposes a multiscale classification network for depth estimation, which tackles the problem by transforming depth values into a nonlinear combination of multiple depth interval values. The network utilizes the correlation of depth information to determine the weight values of each interval. Furthermore, the method employs feature maps with different resolutions to predict depth maps, with lower-resolution maps capturing overall contour and higher-resolution maps providing more accurate object edge details. The desired depth map is obtained through convolution of predicted depth maps at multiple scales.
COMPUTERS & ELECTRICAL ENGINEERING
(2022)
Article
Computer Science, Information Systems
Xin Yang, Qingling Chang, Xinglin Liu, Siyuan He, Yan Cui
Summary: In this paper, a Dense feature fusion network and an adaptive depth fusion module are proposed to fuse multi-scale depth maps effectively for improving the accuracy and object information retrieval in monocular depth estimation. The method enhances the depth map's structure and contour information by integrating diverse depth maps at different scales.
Article
Computer Science, Information Systems
Yifan Zuo, Hao Wang, Yuming Fang, Xiaoshui Huang, Xiwu Shang, Qiang Wu
Summary: This paper proposes a method for enhancing depth maps by introducing guidance from the gradient domain, which improves the quality of depth images and achieves good experimental results.
IEEE TRANSACTIONS ON MULTIMEDIA
(2021)
Article
Automation & Control Systems
Mohammad M. Haji-Esmaeili, Gholamali Montazer
Summary: Estimating the relative depth of a single image is crucial for understanding the structure and relationships within a scene. This paper proposes a new approach that utilizes depth and surface normal datasets from video games, along with a new loss function to improve depth estimation accuracy.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2024)
Article
Engineering, Electrical & Electronic
Xianfa Xu, Zhe Chen, Fuliang Yin
Summary: In this paper, a monocular depth estimation method based on multi-scale feature fusion is proposed, which outperforms existing methods and achieves state-of-the-art results on several public benchmark datasets.
IEEE SIGNAL PROCESSING LETTERS
(2021)
Article
Computer Science, Artificial Intelligence
Xianfa Xu, Zhe Chen, Fuliang Yin
Summary: This study presents a monocular depth estimation method with multi-scale spatial attention guidance and semantic enhancement, which can focus more on small objects and improve the sharpness of depth prediction edges. Experimental results on public benchmark datasets demonstrate the effectiveness and superior performance of the proposed method.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Computer Science, Information Systems
Chenggong Han, Deqiang Cheng, Qiqi Kou, Xiaoyi Wang, Liangliang Chen, Jiamin Zhao
Summary: This paper proposes a self-supervised monocular depth estimation algorithm based on multi-scale structure similarity loss to address the issue of missing depth values in raw depth images. By introducing an attention mechanism to the deep prediction network, the accuracy and perception ability of depth information are further enhanced. Experimental results demonstrate significant improvements in accuracy and visualization effects with the proposed algorithm.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Ali Zia, Jun Zhou, Yongsheng Gao
Summary: This article investigates spectral chromatic and spatial defocus aberration in monocular hyperspectral image (HSI) and proposes methods on how these cues can be utilized for relative depth estimation. By exploring reflectance properties in HSI, three methods are developed for depth estimation. Experimental results show that these methods successfully exploit HSI properties to generate depth cues.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Environmental Sciences
Rong Chang, Kailong Yu, Yang Yang
Summary: Estimating depth from a single low-altitude aerial image captured by a UAS has become a recent research focus. This study proposes a novel multi-scale feature enhancement network and a global scene attention module for depth estimation in low-altitude remote sensing scenes. The results show that our method outperforms state-of-the-art methods on the UAVid 2020 dataset.
Article
Computer Science, Information Systems
Zihao Wang, Sen Yang, Mengji Shi, Kaiyu Qin
Summary: This study proposes a multi-level scale stabilizer (MLSS-VO) combined with a self-supervised feature matching method to address scale uncertainty and scale drift in monocular visual odometry. By extracting feature baselines on different levels and updating motion scale, precise localization and tracking of the target are achieved.
Article
Computer Science, Artificial Intelligence
Zhengfa Liang, Yulan Guo, Yiliu Feng, Wei Chen, Linbo Qiao, Li Zhou, Jianfeng Zhang, Hengzhu Liu
Summary: In this paper, an end-to-end trainable convolutional neural network is proposed for stereo matching, with shared feature extraction, initial disparity estimation, and disparity refinement as its three sub-modules, achieving good performance across multiple datasets.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2021)