Article
Computer Science, Information Systems
In Su Kim, Hyeongbok Kim, Seungwon Lee, Soon Ki Jung
Summary: In this paper, a method for directly predicting the height of objects from a 2D image is proposed. It utilizes an encoder-decoder network for pixel-wise dense prediction based on height consistency. Experimental results demonstrate that the object's height map can be estimated accurately regardless of the camera's location.
Article
Engineering, Civil
Wen Su, Haifeng Zhang, Quan Zhou, Wenzhen Yang, Zengfu Wang
Summary: This paper proposes a method for monocular depth estimation using an information exchange convolutional neural network, effectively integrating local and global context, and improving supervision signals and feature representation.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2021)
Article
Multidisciplinary Sciences
Jan Steinbrener, Vesna Dimitrievska, Federico Pittino, Frans Starmans, Roland Waldner, Jurgen Holzbauer, Thomas Arnold
Summary: We propose a new approach to extract metric volume information of fruits and vegetables from short monocular video sequences and inertial data recorded with a hand-held smartphone. Our approach combines estimated segmentation masks from a pre-trained object detector with predicted change in relative pose obtained from the inertial data to predict the class and volume of the objects of interest. The method achieves high accuracy and comparable results to state-of-the-art setups without requiring reference objects of known size.
Article
Engineering, Electrical & Electronic
Minsoo Song, Seokjae Lim, Wonjun Kim
Summary: A new method for monocular depth estimation is proposed in this paper, which effectively utilizes the Laplacian pyramid in the decoder architecture to improve depth estimation accuracy. Additionally, adjusting weight standardization in convolution blocks can improve gradient flow and make optimization smoother.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2021)
Article
Computer Science, Artificial Intelligence
Qiyu Sun, Yang Tang, Chongzhen Zhang, Chaoqiang Zhao, Feng Qian, Jurgen Kurths
Summary: In this work, the negative impact of dynamic environments on the joint estimation of depth and visual odometry (VO) is alleviated through hybrid masks. The proposed cover mask and filter mask help to improve the accuracy of VO estimation and depth estimation in the presence of dynamic environments.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Engineering, Electrical & Electronic
Seung-Jun Hwang, Sung-Jun Park, Joong-Hwan Baek, Byungkyu Kim
Summary: This article presents a hybrid network for self-supervised monocular depth estimation, which combines CNN and ViT networks. The proposed method achieves higher performance and reduces parameters and computations compared to previous studies, as demonstrated through various experiments.
IEEE SENSORS JOURNAL
(2022)
Review
Chemistry, Analytical
Armin Masoumian, Hatem A. Rashwan, Julian Cristiano, M. Salman Asif, Domenec Puig
Summary: This paper provides a state-of-the-art review of the current developments in monocular depth estimation (MDE) based on deep learning techniques. It highlights the key points from various aspects and discusses limitations and future research directions in the field.
Article
Computer Science, Information Systems
Chuanwu Ling, Xiaogang Zhang, Hua Chen
Summary: This paper proposes an end-to-end unsupervised deep learning framework, which integrates attention blocks and multi-warp loss for monocular depth estimation. Experimental results demonstrate the state-of-the-art performance and satisfactory generalization ability of the proposed method.
IEEE TRANSACTIONS ON MULTIMEDIA
(2022)
Article
Environmental Sciences
Chao Ji, Hong Tang
Summary: This paper proposes three instance-wise gross floor area estimation methods based on monocular optical images. Through experiments and analysis, it is found that there is an inverse relationship between model performance and the degree of end-to-end learning. The proposed methods perform well in gross floor area estimation and provide a new perspective for related tasks.
Article
Geochemistry & Geophysics
Wenbo Sun, Yichen Zhang, Yifan Liao, Biao Yang, Mingchun Lin, Ruifang Zhai, Zhi Gao
Summary: This study reformulates the height estimation task as a classification problem and proposes a unique bin subdivision generation method to improve model performance. Experimental results demonstrate that the proposed method outperforms state-of-the-art approaches in both qualitative and quantitative evaluations.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Engineering, Aerospace
Michele Bechini, Michele Lavagna, Paolo Lunghi
Summary: Recent studies have shown the potential of using monocular images for navigation near uncooperative space objects. However, the development and testing of new algorithms in this field are limited by the availability of spaceborne image datasets. To address this issue, a new algorithm embedded in a tool to generate synthetic high-fidelity spaceborne image datasets is presented. The algorithm is based on open-source ray-tracing software and can be customized for various scenarios.
Article
Computer Science, Artificial Intelligence
Xinchen Ye, Xin Fan, Mingliang Zhang, Rui Xu, Wei Zhong
Summary: By combining Mono-Net and Stereo-Net, this method enhances the performance of monocular depth estimation, enabling the system to better utilize stereo information and achieve superior depth prediction performance.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Computer Science, Artificial Intelligence
Juan Luis Gonzalez Bello, Munchurl Kim
Summary: This paper proposes a novel two-stage training strategy with ambiguity boosting for self-supervised learning of single view depths from stereo images. Experimental results demonstrate the effectiveness and superiority of the proposed method.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Computer Science, Information Systems
Zhen Liang, Tiyu Fang, Yanzhu Hu, Yingjian Wang
Summary: A method for sparse depth densification is proposed in this paper, which utilizes depth densification map and depth error map to exploit the relationship between sparse depth and image for more accurate depth estimation. Experimental results demonstrate the superiority of this approach.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Information Systems
Hyukdoo Choi
Summary: The study explores the potential of monocular depth estimation (MDE) through self-supervised training and novel loss functions to address challenges such as occlusions and dynamic objects, achieving superior results compared to existing unsupervised methods.