Computer vision - ECCV 2022 . Part XXXIII : 17th European Conference, Tel Aviv, Israel, October 23-27, 2022 : proceedings / / Shai Avidan [and four others]

• Pubblicazione: Cham, Switzerland : , : Springer, , [2022]; ©2022
• Descrizione fisica: 1 online resource (804 pages)
• Disciplina: 006.37
• Soggetto topico: Computer vision; Pattern recognition systems
• Persona (resp. second.): AvidanShai
• Nota di contenuto: Intro -- Foreword -- Preface -- Organization -- Contents - Part XXXIII -- SimpleRecon: 3D Reconstruction Without 3D Convolutions -- 1 Introduction -- 2 Related Work -- 2.1 Depth from Calibrated Stereo Pairs -- 2.2 Multi-view Stereo Depth -- 2.3 3D Scene Reconstruction from Posed Views -- 3 Method -- 3.1 Method Overview -- 3.2 Improving the Cost Volume with Metadata -- 3.3 Network Architecture Design -- 3.4 Loss -- 3.5 Implementation Details -- 4 Experiments -- 4.1 Depth Estimation -- 4.2 3D Reconstruction Evaluation -- 4.3 3D Reconstruction Latency -- 4.4 Ablations -- 5 Conclusion -- References -- Structure and Motion from Casual Videos -- 1 Introduction -- 2 Background -- 3 Method -- 3.1 Problem Formulation -- 3.2 Two-Stage Optimization -- 4 Results -- 4.1 Camera Pose and Depth Accuracy on Sintel -- 4.2 Camera Pose Accuracy on TUM Benchmarks -- 4.3 Depth and Pose Quality on DAVIS -- 5 Discussion and Limitations -- References -- What Matters for 3D Scene Flow Network -- 1 Introduction -- 2 Related Work -- 3 3D Scene Flow Network -- 3.1 Network Architecture -- 3.2 Hierarchical Point Feature Abstraction -- 3.3 All-to-All Point Mixture -- 3.4 Hierarchical Flow Refinement -- 4 Experiments -- 4.1 Datasets and Data Preprocess -- 4.2 Training and Evaluation Details -- 5 Results -- 5.1 Comparison with State-of-the-Art (SOTA) -- 5.2 Ablation Study -- 6 Conclusion -- References -- Correspondence Reweighted Translation Averaging -- 1 Introduction -- 2 Related Work -- 2.1 Rotation Averaging -- 2.2 Translation Averaging -- 3 Proposed Method -- 3.1 Our Framework for Translation Averaging -- 4 Experiments -- 4.1 Synthetic Data -- 4.2 Real World Data -- 5 Discussion -- 6 Conclusion -- References -- Neural Strands: Learning Hair Geometry and Appearance from Multi-view Images -- 1 Introduction -- 2 Related Work -- 2.1 Image-Based Hair Modeling. ; 2.2 Neural Hair Rendering -- 3 Overview -- 4 Neural Strands -- 4.1 Neural Scalp Textures -- 4.2 Strand Generator -- 4.3 Neural Hair Renderer -- 5 Training -- 5.1 Strand Generator -- 5.2 End-to-End Optimization -- 6 Results -- 6.1 Evaluation with Synthetic Data -- 6.2 Evaluation with Real Data -- 6.3 Applications -- 7 Limitations and Future Work -- References -- GraphCSPN: Geometry-Aware Depth Completion via Dynamic GCNs -- 1 Introduction -- 2 Related Work -- 3 Method -- 3.1 Spatial Propagation Network -- 3.2 GraphCSPN -- 3.3 Overall Architecture -- 3.4 Loss Function -- 4 Experiments -- 4.1 Datasets and Metrics -- 4.2 Comparison with State-of-the-Arts -- 4.3 Ablation Studies -- 5 Conclusion -- References -- Objects Can Move: 3D Change Detection by Geometric Transformation Consistency -- 1 Introduction -- 2 Related Work -- 3 Detection via Geometric Consistency -- 3.1 Initial Change Detection -- 3.2 Computing Dominant Transformations -- 3.3 Supervoxel Graph Optimization -- 4 Experimental Evaluation -- 5 Conclusion -- References -- Language-Grounded Indoor 3D Semantic Segmentation in the Wild -- 1 Introduction -- 2 Related Work -- 3 Method -- 3.1 Language-Grounded 3D Feature Learning -- 3.2 3D Semantic Segmentation Fine-Tuning -- 3.3 Implementation Details -- 4 ScanNet200 Benchmark -- 5 Experiments -- 6 Conclusion -- References -- Beyond Periodicity: Towards a Unifying Framework for Activations in Coordinate-MLPs -- 1 Introduction -- 2 Related Works -- 3 Methodology -- 3.1 Rank and Memorization -- 3.2 Smoothness and Generalization -- 3.3 Singular Value Distribution as a Proxy for Smoothness -- 3.4 Local Lipschitz Smoothness and the Activation Function -- 4 Experiments -- 4.1 Comparison of Activation Functions -- 4.2 Novel View Synthesis -- 4.3 Convergence -- 4.4 Local Lipschitz Smoothness -- 5 Conclusion -- References. ; Deforming Radiance Fields with Cages -- 1 Introduction -- 2 Related Work -- 3 Method -- 3.1 Radiance Fields Revisited -- 3.2 Cage Generation from Radiance Fields -- 3.3 Cage-Based Deformation -- 3.4 Deforming Radiance Fields -- 4 Implementation Details -- 4.1 Faster Cage Coordinates Computation -- 4.2 Cage Refinement -- 4.3 Radiance Fields Representation -- 5 Experiments -- 5.1 Datasets -- 5.2 Results -- 5.3 Ablation Study -- 5.4 Limitations -- 6 Conclusion -- References -- FLEX: Extrinsic Parameters-free Multi-view 3D Human Motion Reconstruction -- 1 Introduction -- 2 Related Work -- 3 Extrinsic Parameter-free Multi-view Model -- 3.1 Architecture -- 4 Experiments and Evaluation -- 5 Conclusions and Limitations -- References -- MODE: Multi-view Omnidirectional Depth Estimation with 360 Cameras -- 1 Introduction -- 2 Related Work -- 2.1 Deep Learning-Based Stereo Matching Methods -- 2.2 Omnidirectional Depth Estimation -- 2.3 Omnidirectional Depth Datasets -- 3 Multi-view Omnidirectional Depth Estimation -- 3.1 Multi-view Omnidirectional Camera System -- 3.2 Network Architecture -- 3.3 Omnidirectional Stereo Matching with Spherical Convolution -- 3.4 Multi-view Depth Map Fusion -- 4 Datasets -- 5 Experiments -- 5.1 Experimental Settings -- 5.2 Experimental Results -- 5.3 Ablation Studies -- 6 Conclusions -- References -- GigaDepth: Learning Depth from Structured Light with Branching Neural Networks -- 1 Introduction -- 2 Related Work -- 3 Method -- 3.1 Backbone CNN -- 3.2 MLP Tree -- 3.3 Training -- 4 Experiments -- 4.1 Baseline Methods -- 4.2 Dataset -- 4.3 Results -- 4.4 Ablation -- 5 Conclusion -- References -- ActiveNeRF: Learning Where to See with Uncertainty Estimation -- 1 Introduction -- 2 Related Works -- 2.1 Novel View Synthesis -- 2.2 Uncertainty Estimation -- 2.3 Active Learning -- 3 Background -- 4 NeRF with Uncertainty Estimation. ; 5 ActiveNeRF -- 5.1 Prior and Posterior Distribution -- 5.2 Acquisition Function -- 5.3 Optimization and Inference -- 6 Experiments -- 6.1 Experimental Setup -- 6.2 Results -- 7 Conclusion -- References -- PoserNet: Refining Relative Camera Poses Exploiting Object Detections -- 1 Introduction -- 2 Related Work -- 2.1 Relative Pose Estimation -- 2.2 Motion Averaging -- 2.3 Object-Based Computer Vision -- 3 Methodology -- 3.1 Objectness, Matching and Initial Poses -- 3.2 PoserNet: Graph-Based Relative Camera Pose Refinement -- 3.3 Absolute Pose Estimation -- 4 Experiments -- 4.1 Dataset -- 4.2 PoserNet Refining Relative Poses -- 4.3 Absolute Pose Estimates -- 5 Conclusions -- References -- Gaussian Activated Neural Radiance Fields for High Fidelity Reconstruction and Pose Estimation -- 1 Introduction -- 2 Related Work -- 2.1 Neural Scene Representations -- 2.2 Positional Embedding for Pose Estimation -- 2.3 Embedding-Free Coordinate-Networks -- 3 Method -- 3.1 Formulation -- 3.2 Coordinate-Networks -- 3.3 GARF for Reconstruction and Pose Estimation -- 4 Experiments -- 4.1 2D Planar Image Alignment -- 4.2 3D NeRF: Real World Scenes -- 4.3 Implementation Details -- 4.4 Real-World Demo -- 5 Conclusions -- References -- Unbiased Gradient Estimation for Differentiable Surface Splatting via Poisson Sampling -- 1 Introduction -- 2 Related Works -- 3 Statistical Splatting with Unbiased Gradient Estimator -- 3.1 Efficient Pixel-Wise Sampling -- 4 Applications and Results -- 5 Conclusion -- References -- Towards Learning Neural Representations from Shadows -- 1 Introduction -- 1.1 Contributions -- 2 Related Work -- 3 Neural Representations from Shadows -- 3.1 Scenes as Neural Shadow Fields -- 3.2 Differentiable Shadow Mapping -- 3.3 Optimization -- 4 Implementation -- 4.1 Dataset -- 4.2 Training Details -- 5 Results -- 6 Discussion -- 6.1 Future Work. ; 6.2 Conclusions -- References -- Class-Incremental Novel Class Discovery -- 1 Introduction -- 2 Related Works -- 3 Method -- 3.1 Preliminaries -- 3.2 Class-Incremental Novel Class Discovery -- 4 Experiments -- 4.1 Experimental Setup -- 4.2 Ablation Studies -- 4.3 Comparison with State-of-the-Art Methods -- 5 Conclusion -- References -- Unknown-Oriented Learning for Open Set Domain Adaptation -- 1 Introduction -- 2 Related Work -- 2.1 Closed Set Domain Adaptation (CSDA) -- 2.2 Open Set Domain Adaptation (OSDA) -- 3 Unknown-Oriented Learning -- 3.1 True Unknown Excavation -- 3.2 False Unknown Suppression -- 3.3 Known Alignment -- 4 Experiments -- 4.1 Experimental Details -- 4.2 Results for Benchmarks -- 4.3 Ablation Study -- 5 Conclusion -- References -- Prototype-Guided Continual Adaptation for Class-Incremental Unsupervised Domain Adaptation -- 1 Introduction -- 2 Related Work -- 2.1 Unsupervised Domain Adaptation -- 2.2 Class-incremental Domain Adaptation -- 3 Problem Definition -- 4 Prototype-guided Continual Adaptation -- 4.1 Label Prototype Identification -- 4.2 Prototype-Based Alignment and Replay -- 5 Experiments -- 5.1 Experimental Setup -- 5.2 Comparisons with Previous Methods -- 5.3 Application to Enhancing Partial Domain Adaptation -- 5.4 Ablation Studies -- 6 Conclusions -- References -- DecoupleNet: Decoupled Network for Domain Adaptive Semantic Segmentation -- 1 Introduction -- 2 Related Work -- 3 Our Method -- 3.1 Preliminary -- 3.2 Motivation -- 3.3 DecoupleNet -- 3.4 Self-discrimination -- 3.5 Online Enhanced Self-training -- 4 Experiment -- 4.1 Implementation Details -- 4.2 Results -- 4.3 Ablation Study -- 5 Conclusion -- References -- Class-Agnostic Object Counting Robust to Intraclass Diversity -- 1 Introduction -- 2 Related Work -- 3 Analysing Intraclass Diversity for Counting -- 4 Our Robust Class-Agnostic Counter. ; 4.1 Problem Formulation.
• Titolo autorizzato: Computer Vision – ECCV 2022
• ISBN: 3-031-19827-1
• Formato: Materiale a stampa
• Livello bibliografico: Monografia
• Lingua di pubblicazione: Inglese
• Record Nr.: 9910629276903321
• Lo trovi qui: Univ. Federico II
• Serie: Lecture notes in computer science.