Article
Computer Science, Artificial Intelligence
Shangzhe Wu, Christian Rupprecht, Andrea Vedaldi
Summary: We propose a method to learn 3D deformable object categories from raw single-view images without external supervision. The method disentangles the components of the input image using an autoencoder and exploits the underlying object symmetry by reasoning about illumination. Experiments demonstrate its accurate 3D shape recovery for human faces, cat faces, and cars without any supervision.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Software Engineering
Wanquan Feng, Juyong Zhang, Yuanfeng Zhou, Shiqing Xin
Summary: This article addresses the problem of mesh super-resolution and proposes a deep neural network called GDR-Net to solve it. Experimental results demonstrate that GDR-Net outperforms previous methods for recovering geometric details.
IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS
(2022)
Article
Engineering, Electrical & Electronic
Yiman Zhu, Xiao Xiao, Wei Wu, Yu Guo
Summary: In this paper, a cylindrical fitting reconstruction method is proposed for deformable linear objects (DLOs) using a depth camera with only one frame of point clouds. The method preprocesses the point clouds by operation space filtering and outlier removal, segments the specific object using the PointSIFT module, and reconstructs the flexible DLOs using an improved adaptive K-means algorithm. The proposed framework achieves an average error of less than 1 mm during a manipulation experiment in a simulation live-line maintenance site.
SIGNAL IMAGE AND VIDEO PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Shangzhe Wu, Tomas Jakab, Christian Rupprecht, Andrea Vedaldi
Summary: This paper presents DOVE, a method that learns textured 3D models of deformable object categories from monocular videos without keypoint or template supervision. By resolving pose ambiguities and leveraging temporal correspondences, the model automatically factors out 3D shape, articulated pose, and texture from each RGB frame, enabling single-image inference at test time.
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2023)
Article
Computer Science, Artificial Intelligence
Xiang Chen, Nishant Ravikumar, Yan Xia, Rahman Attar, Andres Diaz-Pinto, Stefan K. Piechnik, Stefan Neubauer, Steffen E. Petersen, Alejandro F. Frangi
Summary: In this study, a deep learning architecture called MR-Net is proposed for accurate 3D mesh reconstruction from 2D contours, even with missing data. The method demonstrates superior performance in reconstructing cardiac shapes and shows potential for various applications in medical image analysis.
MEDICAL IMAGE ANALYSIS
(2021)
Article
Robotics
Alessio Caporali, Kevin Galassi, Gianluca Palli
Summary: This letter presents DLO3DS, an approach for estimating and tracking 3D shapes of Deformable Linear Objects (DLOs) using a cheap and compact 2D vision sensor mounted on a robot end-effector. DLO3DS can be applied in various scenarios where perception and manipulation of DLO-like structures are needed. The procedure involves extracting key points from 2D images, modeling DLOs with B-spline curves, and matching the obtained splines using a multi-view stereo-based algorithm. DLO3DS is validated on both real and simulated data, achieving a mean reconstruction error of 0.82 mm on the test set.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Article
Computer Science, Artificial Intelligence
Congying Sui, Kejing He, Congyi Lyu, Yun-Hui Liu
Summary: This paper presents a novel method for 3D reconstruction of dynamic objects, achieving high accuracy and robustness to motion. The method utilizes a structured-light multiplexing technique and extracted codewords to enable high-accuracy encoding, and eliminates reconstruction errors caused by motion between frames. Experimental results validate the high reconstruction accuracy and precision of the proposed method for dynamic objects with different motion speeds and types.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2022)
Article
Geography, Physical
Shengming Yang, Guorong Cai, Jing Du, Ping Chen, Jinhe Su, Yundong Wu, Zongyue Wang, Jonathan Li
Summary: The proposed method automatically reconstructs a building surface model from a raw triangular mesh. It focuses on assemblies of primitives to control the level of geometric detail and incorporates a topological relationship of the primitives to form a plane connection graph.
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2022)
Article
Engineering, Electrical & Electronic
Hui Chen, Fangyong Xu, Wanquan Liu, Dongge Sun, Peter Xiaoping Liu, Muhammad Ilyas Menhas, Bilal Ahmad
Summary: This paper proposes a novel 3D reconstruction method for unstructured objects based on structure from motion in combination with structured light. A new algorithm for point cloud registration and a method for surface reconstruction are introduced to address the challenges in matching point sets from multiple sensors and obtaining smooth surfaces.
IEEE SENSORS JOURNAL
(2021)
Article
Computer Science, Artificial Intelligence
George Fahim, Khalid Amin, Sameh Zarif
Summary: The study proposes a mesh-based single-view object reconstruction model enhanced with additional implicit surface learning, leading to lower surface-to-surface error compared to using an explicit branch alone. The hybrid method outperforms state-of-the-art mesh reconstruction methods and compares favorably to prior work incorporating a hybrid approach. Despite being trained with synthetic images, it generalizes well to real-world images.
IMAGE AND VISION COMPUTING
(2022)
Article
Chemistry, Analytical
Xingyi You, Yue Wang, Xiaohu Zhao
Summary: In the past few years, the 3D Morphing Model (3DMM)-based methods have shown remarkable results in reconstructing 3D faces from a single image. However, generating high-fidelity 3D face texture incurs a computational burden and reduces speed due to the use of deep convolutional neural networks for parameter fitting. To address this issue, we propose Mobile-FaceRNet, an efficient and lightweight network model that combines depthwise separable convolution and multi-scale representation methods for parameter fitting. Experimental results demonstrate that our method achieves high-precision reconstruction while being robust to factors such as pose and occlusion.
Article
Computer Science, Artificial Intelligence
Gianmarco Addari, Jean-Yves Guillemaut
Summary: This paper introduces a family of reconstruction approaches that utilize Helmholtz reciprocity to produce complete 3D models of objects with arbitrary unknown reflectance.
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2023)
Article
Computer Science, Interdisciplinary Applications
B. Aubert, T. Cresson, J. A. de Guise, C. Vazquez
Summary: This paper investigates the robustness and accuracy of intensity-based 3D/2D registration, highlighting the importance of image correspondences. It is found that converting X-ray images into DRR images can improve registration results, especially with the use of GAN-based cross-modality image-to-images translation. The proposed method is applied to precise registration of deformable vertebral models to biplanar radiographs, demonstrating its effectiveness and enhancement.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Article
Robotics
Isabella Huang, Yashraj Narang, Clemens Eppner, Balakumar Sundaralingam, Miles Macklin, Ruzena Bajcsy, Tucker Hermans, Dieter Fox
Summary: Robotic grasping of 3D deformable objects is critical for various real-world applications. This study proposes studying the interaction with deformable objects through physics-based simulation and provides a simulated dataset and code repository for future research. The grasp outcomes on simulated objects show good correspondence with real counterparts.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2022)
Article
Computer Science, Artificial Intelligence
Xianghui Yang, Guosheng Lin, Luping Zhou
Summary: This paper addresses the problem of single-view 3D mesh reconstruction for objects from novel categories not seen during training. It proposes a two-stage network called GenMesh, which breaks the category boundaries by factorizing the image-to-mesh mapping into image-to-point mapping and point-to-mesh mapping. It also introduces a local feature sampling strategy and a multi-view silhouette loss to enhance model generalization and alleviate overfitting.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Stephane Lathuiliere, Benoit Masse, Pablo Mesejo, Radu Horaud
PATTERN RECOGNITION LETTERS
(2019)
Article
Acoustics
Xiaofei Li, Laurent Girin, Sharon Gannot, Radu Horaud
IEEE-ACM TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING
(2019)
Article
Engineering, Electrical & Electronic
Xiaofei Li, Yutong Ban, Laurent Girin, Xavier Alameda-Pineda, Radu Horaud
IEEE JOURNAL OF SELECTED TOPICS IN SIGNAL PROCESSING
(2019)
Article
Engineering, Electrical & Electronic
Yutong Ban, Xavier Alameda-Pineda, Christine Evers, Radu Horaud
IEEE SIGNAL PROCESSING LETTERS
(2019)
Article
Engineering, Electrical & Electronic
Xiaofei Li, Simon Leglaive, Laurent Girin, Radu Horaud
IEEE SIGNAL PROCESSING LETTERS
(2019)
Article
Acoustics
Xiaofei Li, Laurent Girin, Sharon Gannot, Radu Horaud
IEEE-ACM TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING
(2019)
Article
Computer Science, Artificial Intelligence
Stephane Lathuiliere, Pablo Mesejo, Xavier Alameda-Pineda, Radu Horaud
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2020)
Article
Computer Science, Artificial Intelligence
Yutong Ban, Xavier Alameda-Pineda, Laurent Girin, Radu Horaud
Summary: This article introduces a generative audio-visual fusion model for tracking multiple speakers and demonstrates its effectiveness in informal meetings through experiments.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2021)
Proceedings Paper
Acoustics
Simon Leglaive, Xavier Alameda-Pineda, Laurent Girin, Radu Horaud
2020 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH, AND SIGNAL PROCESSING
(2020)
Article
Acoustics
Mostafa Sadeghi, Simon Leglaive, Xavier Alameda-Pineda, Laurent Girin, Radu Horaud
IEEE-ACM TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING
(2020)
Proceedings Paper
Acoustics
Xiaofei Li, Radu Horaud
2019 IEEE WORKSHOP ON APPLICATIONS OF SIGNAL PROCESSING TO AUDIO AND ACOUSTICS (WASPAA)
(2019)
Proceedings Paper
Computer Science, Interdisciplinary Applications
Xavier Alameda-Pineda, Soraya Arias, Yutong Ban, Guillaume Delorme, Laurent Girin, Radu Horaud, Xiaofei Li, Bastien Mourgue, Guillaume Sarrazin
PROCEEDINGS OF THE 27TH ACM INTERNATIONAL CONFERENCE ON MULTIMEDIA (MM'19)
(2019)
Proceedings Paper
Engineering, Electrical & Electronic
Benoit Masse, Stephane Lathuiliere, Pablo Mesejo, Radu Horaud
2019 14TH IEEE INTERNATIONAL CONFERENCE ON AUTOMATIC FACE AND GESTURE RECOGNITION (FG 2019)
(2019)
Proceedings Paper
Acoustics
Simon Leglaive, Laurent Girin, Radu Horaud
2019 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP)
(2019)
Article
Computer Science, Artificial Intelligence
Xiaofei Li, Laurent Girin, Radu Horaud
CAAI TRANSACTIONS ON INTELLIGENCE TECHNOLOGY
(2019)