Pattern recognition and computer vision . Part II : 4th Chinese Conference, PRCV 2021, Beijing, China, October 29-November 1, 2021, Proceedings / / Huimin Ma [and seven others] (editors)

• Pubblicazione: Cham, Switzerland : , : Springer, , [2021]; ©2021
• Descrizione fisica: 1 online resource (694 pages)
• Disciplina: 006.4
• Soggetto topico: Pattern recognition systems; Computer vision
• Persona (resp. second.): MaHuimin
• Nota di bibliografia: Includes bibliographical references and index.
• Nota di contenuto: Intro -- Preface -- Organization -- Contents - Part II -- Computer Vision, Theories and Applications -- Dynamic Fusion Network for Light Field Depth Estimation -- 1 Introduction -- 2 Related Work -- 3 Method -- 3.1 The Overall Architecture -- 3.2 Pyramid ConvGRU -- 3.3 Multi-modal Dynamic Fusion Module (MDFM) -- 4 Experiments -- 4.1 Experiments Setup -- 4.2 Ablation Studies -- 4.3 Comparison with State-of-the-arts -- 5 Conclusion -- References -- Metric Calibration of Aerial On-Board Multiple Non-overlapping Cameras Based on Visual and Inertial Measurement Data -- 1 Introduction -- 2 Related Works -- 3 Metric Calibration Based on Visual and Inertial Measurement Data -- 3.1 Notation and Problem Formulation -- 3.2 Relative Pose Estimation via Structure from Motion -- 3.3 Inertial Measurement Data Based Metric Scale Factor Estimation -- 4 Experimental Results -- 4.1 Equipment -- 4.2 Metric Calibration of the Aerial On-Board Non-overlapping Camera System -- 4.3 Metric Calibration of an Industrial Non-overlapping Camera System -- 4.4 Experiments of Applications for Object Metric 3D Reconstruction -- 5 Conclusions -- References -- SEINet: Semantic-Edge Interaction Network for Image Manipulation Localization -- 1 Introduction -- 2 Related Work -- 3 Method -- 3.1 Cross Interaction Pattern -- 3.2 Aggregate Interaction Module -- 3.3 Bidirectional Fusion Module -- 3.4 Training Loss -- 4 Experiments -- 4.1 Datasets and Implementation Details -- 4.2 Evaluation Metrics -- 4.3 Ablation Studies -- 4.4 Robustness Analysis -- 4.5 Comparing with State-of-the-Art -- 5 Conclusion -- References -- Video-Based Reconstruction of Smooth 3D Human Body Motion -- 1 Introduction -- 2 Related Work -- 2.1 3D Human Mesh from Single Images -- 2.2 3D Human Mesh from Video -- 2.3 GANs for Modeling -- 3 Approach -- 3.1 3D Body Representation -- 3.2 Temporal Encoder. ; 3.3 Constraint Loss -- 3.4 Motion Discriminator -- 4 Experiments -- 4.1 Implement Details -- 4.2 Comparison to Other Methods -- 4.3 Ablation Experiments -- 5 Conclusion -- References -- A Unified Modular Framework with Deep Graph Convolutional Networks for Multi-label Image Recognition -- 1 Introduction -- 2 Related Work -- 3 Proposed Method -- 3.1 Image Feature Extraction Module -- 3.2 Label Semantic Extraction Module -- 3.3 Prediction Results and Training Scheme -- 4 Experiments -- 4.1 Evaluation Metrics -- 4.2 Implementation Details -- 4.3 Experimental Results -- 4.4 Ablation Studies -- 4.5 Adjacency Matrix Visualization -- 5 Conclusion -- References -- 3D Correspondence Grouping with Compatibility Features -- 1 Introduction -- 2 Related Work -- 2.1 3D Correspondence Grouping -- 2.2 Learning for Correspondence Grouping -- 3 Methodology -- 3.1 Compatibility Check -- 3.2 CF Feature Extraction -- 3.3 CF Classification -- 4 Experiments -- 4.1 Experimental Setup -- 4.2 Method Analysis -- 4.3 Comparative Results and Visualization -- 5 Conclusions -- References -- Contour-Aware Panoptic Segmentation Network -- 1 Introduction -- 2 Related Work -- 3 Approach -- 3.1 Panoptic Contour Branch -- 3.2 Panoptic Segmentation Branch -- 3.3 Structure Loss Function -- 4 Experiments -- 4.1 Dataset -- 4.2 Evaluation Metrics -- 4.3 Implementation Details -- 4.4 Comparisons with Other Methods -- 4.5 Ablative Analysis -- 5 Conclusion -- References -- VGG-CAE: Unsupervised Visual Place Recognition Using VGG16-Based Convolutional Autoencoder -- 1 Introduction -- 2 Realted Work -- 2.1 Handcraft-Based Methods -- 2.2 CNN-Based Methods -- 2.3 AE-Based Methods -- 3 VGG16-Based Convolutional Autoencoder -- 3.1 Model Architecture -- 3.2 Training -- 3.3 Matching -- 4 Experiments -- 4.1 Datasets -- 4.2 State-of-the-Art Approaches -- 4.3 Ground Truth -- 4.4 Comparison and Discussion. ; 5 Conclusion -- References -- Slice Sequential Network: A Lightweight Unsupervised Point Cloud Completion Network -- 1 Introduction -- 2 Related Work -- 2.1 3D Learning -- 2.2 3D Completion -- 3 Our Method -- 3.1 Overview -- 3.2 Slicer -- 3.3 Multi-scale Point Encoder -- 3.4 Sequential Predictor -- 3.5 Shape Prediction Decoder -- 3.6 Loss Function -- 4 Experiments -- 4.1 Datasets and Implementation Details -- 4.2 Point Cloud Completion Results -- 4.3 Analysis of Encoder -- 4.4 Robustness to Occlusion -- 4.5 Comparison of Complexity -- 5 Ablation Study -- 6 Conclusion -- References -- From Digital Model to Reality Application: A Domain Adaptation Method for Rail Defect Detection -- 1 Introduction -- 2 Preliminaries -- 3 Method -- 3.1 DT-Based Virtual Data Generation -- 3.2 Dummy-Target Domain -- 3.3 DA-YOLO -- 4 Experiment -- 4.1 Dataset and Evaluation Metrics -- 4.2 Experiment Settings -- 4.3 Experimental Results -- 5 Conclusion -- References -- FMixAugment for Semi-supervised Learning with Consistency Regularization -- 1 Introduction -- 2 Related Work -- 3 Methods -- 3.1 FMixAugment: MixAugment Combined with FMask -- 3.2 Improved Consistency Regularization -- 3.3 Dynamic Growth Threshold -- 4 Experiments -- 4.1 Implementation Details -- 4.2 Experimental Results -- 4.3 Ablation Study -- 5 Conclusion and Future Work -- References -- IDANet: Iterative D-LinkNets with Attention for Road Extraction from High-Resolution Satellite Imagery -- 1 Introduction -- 2 Related Work -- 3 Methodology -- 3.1 Overview -- 3.2 Basic Iteration Module -- 3.3 Iterative Architecture -- 4 Experiment -- 4.1 Datasets -- 4.2 Implementation Details -- 5 Results -- 5.1 Comparison of Road Segmentation Methods -- 5.2 Ablation Experiment -- 5.3 The Influence of Network Iteration -- 6 Conclusion -- References. ; Disentangling Deep Network for Reconstructing 3D Object Shapes from Single 2D Images -- 1 Introduction -- 2 Related Works -- 3 Disentangling Deep Network -- 3.1 Network Architecture -- 3.2 Learning Objective Functions -- 3.3 Training Strategy -- 4 Experiments -- 4.1 Implementation Details -- 4.2 Ablation Analysis -- 4.3 3D Reconstruction -- 4.4 Effects of 3D Shape Identity -- 5 Conclusion -- References -- AnchorConv: Anchor Convolution for Point Clouds Analysis -- 1 Introduction -- 2 Related Work -- 3 Proposed Method -- 3.1 AnchorConv -- 3.2 Anchor Reweighting Module -- 3.3 Network Architectures -- 4 Experiments -- 4.1 Classification on ModelNet40 -- 4.2 ShapeNet Part Segmentation -- 4.3 3D Segmentation of Indoor Scene -- 4.4 3D Segmentation of Outdoor Scene -- 4.5 Ablation Study -- 4.6 Qualitative Results -- 5 Conclusion -- References -- IFR: Iterative Fusion Based Recognizer for Low Quality Scene Text Recognition -- 1 Introduction -- 2 Related Works -- 3 Method -- 3.1 Iterative Collaboration -- 3.2 Fusion Module RRF -- 3.3 Loss Functions -- 3.4 Paired Training Data Generate -- 4 Experiments -- 4.1 Datasets and Implementation Details -- 4.2 Ablation Study -- 4.3 Comparisons with State-of-the-Arts -- 5 Conclusion -- References -- Immersive Traditional Chinese Portrait Painting: Research on Style Transfer and Face Replacement -- 1 Introduction -- 2 Related Work -- 2.1 Neural Style Transfer -- 2.2 Face Replacement -- 3 The P-CP Method -- 3.1 Network Architecture -- 3.2 Neural Style Transfer Network -- 3.3 Face Replacement -- 4 Experiment -- 4.1 Comparison of Different Traditional Chinese Painting Styles -- 4.2 Image Detail Exploration and Optimization -- 4.3 Improvement of Face Replacement with Style Transfer -- 5 Conclusion -- References -- Multi-camera Extrinsic Auto-calibration Using Pedestrians in Occluded Environments -- 1 Introduction. ; 2 Related Work -- 3 Calibration Based on 3D Positions -- 3.1 3D Head Positions in Local Camera Coordinates -- 3.2 Registration of 3D Point Sets -- 4 Refinement -- 5 Experiments and Results -- 6 Conclusion -- References -- Dual-Layer Barcodes -- 1 Introduction -- 2 Related Work -- 2.1 Steganography -- 2.2 Watermarking -- 2.3 Barcode -- 3 Method -- 3.1 Encoder -- 3.2 Decoder -- 3.3 Noise Layer -- 3.4 Discriminator -- 4 Experiments and Analysis -- 4.1 Dataset and Experimental Setting -- 4.2 Implementation Details -- 4.3 Metrics -- 5 Discussion -- 6 Conclusion -- References -- Graph Matching Based Robust Line Segment Correspondence for Active Camera Relocalization -- 1 Introduction -- 2 Method -- 2.1 System Overview -- 2.2 Robust Line Segment Matching -- 2.3 Active Camera Relocation -- 3 Experiments -- 3.1 Experimental Setup -- 3.2 Analysis of Line Segment Matching -- 3.3 Analysis of Relocalization Accuracy and Convergence Speed -- 3.4 Analysis of Robustness in Hard Scenes -- 4 Conclusion -- References -- Unsupervised Learning Framework for 3D Reconstruction from Face Sketch -- 1 Introduction -- 2 Related Work -- 2.1 Image-to-Image Translation -- 2.2 3D Shape Reconstruction -- 3 Method -- 3.1 Dataset Construction -- 3.2 Network Architecture -- 3.3 Loss Functions -- 4 Experiments -- 4.1 Implementation Details -- 4.2 Quantitative Results and Ablation Study -- 4.3 Qualitative Results -- 5 Conclusion -- References -- HEI-Human: A Hybrid Explicit and Implicit Method for Single-View 3D Clothed Human Reconstruction -- 1 Introduction -- 2 Related Work -- 3 Methodology -- 3.1 Overview -- 3.2 Explicit Model -- 3.3 Implicit Model -- 3.4 Loss Functions -- 4 Experiments -- 4.1 Dataset and Protocol -- 4.2 Training Details -- 4.3 Quantitative Results -- 4.4 Qualitative Results -- 4.5 Ablation Studies -- 5 Conclusions -- References. ; A Point Cloud Generative Model via Tree-Structured Graph Convolutions for 3D Brain Shape Reconstruction.
• Titolo autorizzato: Pattern recognition and computer vision
• ISBN: 3-030-88007-9
• Formato: Materiale a stampa
• Livello bibliografico: Monografia
• Lingua di pubblicazione: Inglese
• Record Nr.: 996464410803316
• Lo trovi qui: Univ. di Salerno
• Serie: Lecture notes in computer science ; ; 13020.