Article
Computer Science, Artificial Intelligence
Bin Chen, Jian Zhang
Summary: CASNet is a novel content-aware scalable network that efficiently addresses image compressed sensing (CS) problems by achieving adaptive sampling rate allocation and high-quality reconstruction. It utilizes a data-driven saliency detector to evaluate the importance of different image regions and a saliency-based strategy for sampling rate allocation. By using a unified learnable generating matrix and an optimization-inspired recovery subnet, CASNet jointly reconstructs image blocks sampled at various sampling rates. The proposed SVD-based initialization and random transformation enhancement strategies improve training convergence and network robustness. Experiments show that CASNet outperforms other CS networks, demonstrating the collaboration and mutual supports among its components and strategies.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2022)
Article
Computer Science, Artificial Intelligence
Miguel Domingo, Francisco Casacuberta
Summary: Historical documents are crucial to our cultural heritage, but modernizing their language and achieving spelling consistency is a challenging task. This study proposes an interactive framework based on machine translation to aid scholars in generating error-free modernizations/normalizations. The evaluation conducted in simulated environments shows significant reductions in human effort.
PATTERN ANALYSIS AND APPLICATIONS
(2023)
Article
Geochemistry & Geophysics
Xueyao Xiao, Wei Zhang, Yi Chang, Shuning Cao, Wei He, Houzhang Fang, Luxin Yan
Summary: The study proposes a content-aware subspace low-rank tensor recovery method, utilizing deep network and low-rank tensor model to adaptively learn the optimal subspace dimension for improved restoration of hyperspectral images.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Computer Science, Artificial Intelligence
Ivo Pisacovic, Frantisek Darena, David Prochazka, Vit Janis
Summary: Establishing normative documents is essential for larger organizations to control processes and provide solutions to common problems, but the formal and difficult-to-read nature of these documents necessitates different customer services. Companies are increasingly developing chatbots for firstline customer support automation, but automatic answering directly from normative documents is often ineffective. A novel preprocessing method is proposed in this paper to improve the accuracy of automatic question answering on normative documents.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Geochemistry & Geophysics
Reuma Arav, Sagi Filin, Norbert Pfeifer
Summary: In this article, a context-aware subsampling approach is proposed to reduce the data load of less important regions while retaining high resolution of objects of interest. Visual saliency measures are used to identify regions that require detail preservation, and data reduction is only applied to nonsalient regions. A hierarchical data structure based on surface nature enables progressive subsampling, with retained representative points describing the underlying surface. The proposed model is demonstrated on datasets from different scanners, and results are compared with three common simplification approaches, showing a reduced point cloud similar to the original for ROI analysis regardless of the level of simplification.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Computer Science, Artificial Intelligence
Luca Calatroni, Martin Huska, Serena Morigi, Giuseppe Antonio Recupero
Summary: In recent years, there has been significant development in technologies for 3D data acquisition and 3D printing. This has led to an increased demand for 3D virtual models of scanned objects. This study presents a geometric framework that integrates feature-aware denoising, hole filling, and context-aware completion to recover triangulated surfaces from damaged and incomplete noisy observations. The proposed framework, based on a unified variational model, shows robustness and elegance in accurate restorations even in the presence of severe random noise and large damaged areas.
JOURNAL OF MATHEMATICAL IMAGING AND VISION
(2023)
Article
Computer Science, Artificial Intelligence
Jingwen He, Chao Dong, Yihao Liu, Yu Qiao
Summary: The paper introduces a new multi-dimension modulation method to adjust output effects across multiple degradation types and levels, which is more effective in handling various degradation types and alleviating data imbalance compared to previous single-dimension modulation.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Computer Science, Artificial Intelligence
Mantang Guo, Junhui Hou, Jing Jin, Hui Liu, Huanqiang Zeng, Jiwen Lu
Summary: This paper proposes a content-aware warping method that adaptsively learns the interpolation weights for pixels from their contextual information via a lightweight neural network. Based on this learnable warping module, a new end-to-end learning-based framework is proposed for novel view synthesis, which includes two additional modules to address occlusion and spatial correlation issues. Experimental results demonstrate that the proposed method outperforms state-of-the-art methods both quantitatively and visually.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Jun Liu, Ming Yan, Tieyong Zeng
Summary: Blind image deblurring is a difficult problem that requires proper prior knowledge. The surface-aware strategy proposed in this paper facilitates blur kernel estimation and outperforms other methods in experiments.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2021)
Article
Computer Science, Information Systems
Yue Zhang, Chao Liang, Longxiang Jiang
Summary: This study proposes a confidence-aware active feedback method (CAAF) specifically designed for online relevance feedback in interactive instance search tasks. By utilizing a pairwise manifold ranking loss to evaluate the ranking confidence of each unlabeled sample, CAAF improves both the interaction efficiency and ranking quality. In addition, two acceleration strategies, an approximate optimization scheme and a top-K search scheme, are designed to reduce the computational complexity of CAAF. Extensive experiments demonstrate the effectiveness of the proposed method in both image and video instance search tasks, particularly in the large-scale video instance search task of NIST TRECVID 2021.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Article
Chemistry, Multidisciplinary
Yago Diez, Toya Suzuki, Marius Vila, Katsushi Waki
Summary: This paper introduces an automatic method for detecting and classifying kanji characters within historical Japanese Wasan documents using deep learning. By exploring the use of modern kanji databases for classification, new possibilities are discovered. Experimental results show high accuracy and low false detection rates.
APPLIED SCIENCES-BASEL
(2021)
Article
Computer Science, Software Engineering
George E. Brown, Rahul Narain
Summary: This paper proposes two improvements, a local-global splitting based on the polar decomposition and a dynamic reweighting method, to address the challenges of local-global solvers in handling large rotations and deformations. The improved algorithm outperforms state-of-the-art approaches in quasi-static simulations and parameterization problems.
ACM TRANSACTIONS ON GRAPHICS
(2021)
Article
Computer Science, Information Systems
Pegah Alizadeh, Aomar Osmani, Mohamed Essaid Khanouche, Abdelghani Chibani, Yacine Amirat
Summary: The paper addresses the challenge of selecting an optimal service composition without prior knowledge of user QoS preferences, proposing a vector-valued MDP approach that successfully finds the optimal composite services with around 50 interactions with the user.
IEEE SYSTEMS JOURNAL
(2021)
Article
Computer Science, Software Engineering
Shahid Latif, Zheng Zhou, Yoon Kim, Fabian Beck, Nam Wook Kim
Summary: Charts and text are commonly used together to convey complex data in news articles, online blogs, and academic papers. The tight coupling of text and charts in data documents is challenging, as it requires the synthesis of information across the two modalities. This research explores ways to support this tight coupling and develops a mixed-initiative interface that enables users to construct interactive references between text and charts. The results of user studies and algorithmic evaluations indicate that the interface provides an effective way to compose interactive data documents.
IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS
(2022)
Article
Computer Science, Software Engineering
Rui Wang, Zhihao Zheng, Weishan Yu, Yanli Shao, Shuming Gao
Summary: The proposed structure-aware geometric optimization method for hex meshes improves the overall parameterized energy by relocating the base complex position, resulting in optimized geometric embedding.
COMPUTER-AIDED DESIGN
(2021)
Article
Computer Science, Software Engineering
M. Kluge, T. Weyrich, A. Kolb
COMPUTER GRAPHICS FORUM
(2020)
Article
Computer Science, Artificial Intelligence
Carlo Innamorati, Tobias Ritschel, Tim Weyrich, Niloy J. Mitra
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2020)
Article
Humanities, Multidisciplinary
Karina Rodriguez Echavarria, Myrsini Samaroudi, Tim Weyrich
ACM JOURNAL ON COMPUTING AND CULTURAL HERITAGE
(2020)
Article
Computer Science, Software Engineering
Gilles Rainer, Abhijeet Ghosh, Wenzel Jakob, Tim Weyrich
COMPUTER GRAPHICS FORUM
(2020)
Article
Optics
Oskar Elek, Ran Zhang, Denis Sumin, Karol Myszkowski, Bernd Bickel, Alexander Wilkie, Jaroslav Krivanek, Tim Weyrich
Summary: This paper presents a novel measurement approach for indirectly inferring transport parameters of volumetric light from extrinsic observations, enabling accurate simulations of complex materials and precise reproduction of appearance in 3D printing. The method can potentially be used for obtaining spectral measurements for other application areas as well, without fundamental changes to the basic measurement methodology.
Article
Computer Science, Software Engineering
Jose Ezequiel Soto Sanchez, Tim Weyrich, Asla Medeiros e Sa, Luiz Henrique de Figueiredo
Summary: The paper introduces a representation method for periodic tilings of the plane using regular polygons, where all elements of the tiling are systematically generated from a small subset of seed vertices by translations. The concrete representation of a tiling is done using a (2 + n) x4 integer matrix containing lattice coordinates for two translation vectors and n seed vertices. Various properties of this representation are discussed, along with how to efficiently exploit it for reconstruction, rendering, and automatic crystallographic classification by symmetry detection.
COMPUTERS & GRAPHICS-UK
(2021)
Article
Computer Science, Software Engineering
B. R. Mallikarjun, Ayush Tewari, Abdallah Dib, Tim Weyrich, Bernd Bickel, Hans-Peter Seidel, Hanspeter Pfister, Wojciech Matusik, Louis Chevallier, Mohamed Elgharib, Christian Theobalt
Summary: This paper presents a method that learns from limited supervised training data for high-quality intuitive editing of camera viewpoint and scene illumination in head portraits; by operating in the generative model space of StyleGAN, combining the best of supervised learning and generative adversarial modeling, it achieves high-quality photorealistic editing of in-the-wild images.
ACM TRANSACTIONS ON GRAPHICS
(2021)
Article
Computer Science, Software Engineering
Alejandro Sztrajman, Gilles Rainer, Tobias Ritschel, Tim Weyrich
Summary: The article introduces a compact neural network-based representation of BRDF data that combines high-accuracy reconstruction with efficient practical rendering. Encoding BRDFs as lightweight networks and proposing a training scheme with adaptive angular sampling are critical for accurate reconstruction of specular highlights. Additionally, a novel approach is proposed to make the representation adaptable to importance sampling.
COMPUTER GRAPHICS FORUM
(2021)
Review
Chemistry, Applied
David Zargaran, Florence Zoller, Alexander Zargaran, Tim Weyrich, Afshin Mosahebi
Summary: This study provides a comprehensive overview of the pathways and mechanisms underlying facial ageing, including both intrinsic and extrinsic factors. These alterations in pathways result in negative changes in the skin such as wrinkles, increased laxity, and thinning. Understanding these mechanisms can help inform further research on slowing down or impeding the ageing process and developing new treatment strategies.
INTERNATIONAL JOURNAL OF COSMETIC SCIENCE
(2022)
Article
Humanities, Multidisciplinary
Charlie Willard, Nancy Wade, Matija Strli, John R. Gilchrist, Tim Weyrich, Adam Gibson
Summary: This paper introduces two approaches to correct dropped frames in line-scan cameras using the A* search algorithm. By comparing the two methods, it is found that aligning overlapping sections of images performs better than aligning to a reference image. These methods are then applied to mosaic high-resolution hyperspectral images of a cultural heritage painting, resulting in a composite image with high spatial and spectral resolution.
ACM JOURNAL ON COMPUTING AND CULTURAL HERITAGE
(2022)
Article
Surgery
David Zargaran, Alexander Zargaran, Tom Terranova, Helia Khaledi, Alexandra Robinson, Julie Davies, Tim Weyrich, Afshin Mosahebi
Summary: This paper provides a descriptive and qualitative analysis of the advertised practitioners in the injectables market in the UK. It identifies a diverse range of practitioners with different professional backgrounds and experiences, highlighting the potential impact on patient risks and the importance of introducing licensing to the industry.
JOURNAL OF PLASTIC RECONSTRUCTIVE AND AESTHETIC SURGERY
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
David R. Walton, Koray Kavakli, Rafael Kuffner dos Anjos, David Swapp, Tim Weyrich, Hakan Urey, Anthony Steed, Tobias Ritschel, Kaan Aksit
Summary: In the study of computer-generated holography, we propose a new method that utilizes gaze-contingency and perceptual graphics to accelerate the development of practical holographic display systems. By inferring the user's focal depth and generating images only at their focus plane, we are able to improve foveal visual quality and reduce distortions in peripheral vision. We also introduce a novel metameric loss function for comparing image statistics and implement a model representing the relation between holograms and image reconstructions.
2022 IEEE CONFERENCE ON VIRTUAL REALITY AND 3D USER INTERFACES (VR 2022)
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
B. R. Mallikarjun, Ayush Tewari, Tae-Hyun Oh, Tim Weyrich, Bernd Bickel, Hans-Peter Seidel, Hanspeter Pfister, Wojciech Matusik, Mohamed Elgharib, Christian Theobalt
Summary: The study introduces a new neural representation for describing the reflectance properties of faces, enabling estimation of all reflectance components from a monocular image for better face rendering results.
2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021
(2021)
Article
Computer Science, Software Engineering
Tobias Rittig, Denis Sumin, Vahid Babaei, Piotr Didyk, Alexey Voloboy, Alexander Wilkie, Bernd Bickel, Karol Myszkowski, Tim Weyrich, Jaroslav Krivanek
Summary: With the rise of full-color 3D printers, the need for color-accurate 3D-print preparation has increased, but traditional printing materials' translucency poses a challenge in maintaining color texture details. Our data-driven approach significantly speeds up the optimization process, achieving results similar in quality to Monte Carlo rendering but in a fraction of the time.
COMPUTER GRAPHICS FORUM
(2021)
Proceedings Paper
Engineering, Electrical & Electronic
Alejandro Sztrajman, Alexandros Neophytou, Tim Weyrich, Eric Sommerlade
2020 INTERNATIONAL CONFERENCE ON 3D VISION (3DV 2020)
(2020)