Computer vision - ECCV 2022 . Part XV : 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 (799 pages)
• Disciplina: 006.4
• Soggetto topico: Pattern recognition systems; Computer vision
• Persona (resp. second.): AvidanShai
• Nota di contenuto: Intro -- Foreword -- Preface -- Organization -- Contents - Part XV -- WaveGAN: Frequency-Aware GAN for High-Fidelity Few-Shot Image Generation -- 1 Introduction -- 2 Related Work -- 3 Methodology -- 3.1 Overview -- 3.2 WaveEncoder -- 3.3 WaveDecoder -- 3.4 Optimization Objective -- 4 Experiments -- 4.1 Quantitative Evaluation -- 4.2 Qualitative Evaluation -- 4.3 Visualization of the Frequency Components of Generated Images -- 4.4 Ablation Studies -- 4.5 Augmentation for Classification -- 4.6 Influence of the Number of Shots -- 5 Conclusion and Acknowledgments -- References -- End-to-End Visual Editing with a Generatively Pre-trained Artist -- 1 Introduction -- 2 Related Work -- 3 Method -- 3.1 End-to-End Conditional Generation -- 3.2 Synthesizing a Dataset of Meaningful Edits -- 3.3 Two-Stage Conditional Auto-regressive Image Generation -- 4 Experiments -- 4.1 Quantitative Evaluation -- 4.2 Qualitative Evaluation -- 4.3 Ablations -- 5 Conclusions, Limitations and Future Work -- References -- High-Fidelity GAN Inversion with Padding Space -- 1 Introduction -- 2 Related Work -- 3 Method -- 3.1 Preliminaries -- 3.2 Padding Space for GAN Inversion -- 3.3 Encoder Architecture -- 3.4 Training Objectives -- 4 Experiments -- 4.1 Experimental Settings -- 4.2 GAN Inversion Performance -- 4.3 Separate Control of Spatial Contents and Image Style -- 4.4 Manipulation with Customized Image Pair -- 5 Conclusion -- References -- Designing One Unified Framework for High-Fidelity Face Reenactment and Swapping -- 1 Introduction -- 2 Related Work -- 3 Method -- 3.1 Architecture -- 3.2 Face Reenactment and Swapping -- 3.3 Objective Functions -- 4 Experiments -- 4.1 Datasets and Implementations Details -- 4.2 Comparison with SOTAs -- 4.3 Ablation Study and Further Analysis -- 5 Conclusions -- References. ; Sobolev Training for Implicit Neural Representations with Approximated Image Derivatives -- 1 Introduction -- 2 Related Work -- 2.1 Implicit Neural Representations -- 2.2 Derivative Supervision -- 3 Methodology -- 3.1 Formulation -- 3.2 Approximate Derivatives with Finite Differences -- 3.3 Activation Functions -- 4 Experiments -- 4.1 Direct: Image Regression -- 4.2 Indirect: Inverse Rendering -- 4.3 Ablation Study -- 5 Discussions -- 5.1 Future Work -- 5.2 Limitations -- 6 Conclusion -- References -- .26em plus .1em minus .1emMake-A-Scene: Scene-Based Text-to-Image Generation with Human Priors -- 1 Introduction -- 2 Related Work -- 3 Method -- 3.1 Scene Representation and Tokenization -- 3.2 Adhering to Human Emphasis in the Token Space -- 3.3 Scene-Based Transformer -- 3.4 Autoregressive Transformer Classifier-Free Guidance -- 4 Experiments -- 4.1 Comparison with Previous Work -- 4.2 Scene Controllability -- 4.3 Ablation Study -- 5 Conclusion -- References -- 3D-FM GAN: Towards 3D-Controllable Face Manipulation -- 1 Introduction -- 2 Related Works -- 3 Methodology -- 3.1 Overview and Notations -- 3.2 Dataset -- 3.3 Training Strategy -- 3.4 Architecture -- 4 Experiment -- 4.1 Experimental Setup -- 4.2 Evaluation Metrics -- 4.3 Architectures Evaluation -- 4.4 Controllable Image Synthesis -- 5 Comparison to State of the Arts -- 5.1 Quantitative Comparison -- 5.2 Visual Comparison -- 6 Ablation Study -- 6.1 Training Strategy -- 6.2 Two-Phase Training -- 7 Conclusion -- References -- Multi-Curve Translator for High-Resolution Photorealistic Image Translation -- 1 Introduction -- 2 Methodology -- 2.1 Problem Formulation and Prior Work -- 2.2 Multi-Curve Translator -- 2.3 Training Strategy -- 3 Applications -- 3.1 Photorealistic I2I Translation -- 3.2 Style Transfer -- 3.3 Image Dehazing -- 3.4 Photo Retouching -- 4 Experiments -- 4.1 Runtime. ; 4.2 Qualitative Comparison -- 4.3 Quantitative Comparison -- 5 Discussion -- References -- Deep Bayesian Video Frame Interpolation -- 1 Introduction -- 2 Related Works -- 3 The Proposed Method -- 3.1 Background and Notations -- 3.2 Deep Bayesian Video Frame Interpolation -- 3.3 Formulating the Gradients -- 3.4 Implementation -- 4 Experiments -- 4.1 Experimental Setup -- 4.2 Comparisons with State-of-the-Art Models -- 4.3 Performance Analysis -- 5 Conclusion -- References -- Cross Attention Based Style Distribution for Controllable Person Image Synthesis -- 1 Introduction -- 2 Related Work -- 3 Method -- 3.1 Overview Framework -- 3.2 Pre- and Post-attention Style Injection -- 3.3 Cross Attention Based Style Distribution Block -- 3.4 Learning Objectives -- 4 Experiments -- 4.1 Experimental Setup -- 4.2 Pose Transfer -- 4.3 Ablation Study -- 4.4 Virtual Try-On and Head Swapping -- 4.5 Target Parsing Map Synthesis -- 5 Limitations -- 6 Conclusion -- References -- KeypointNeRF: Generalizing Image-Based Volumetric Avatars Using Relative Spatial Encoding of Keypoints -- 1 Introduction -- 2 Related Work -- 3 Preliminaries: Neural Radiance Fields -- 4 KeypointNeRF -- 4.1 Relative Spatial Keypoint Encoding -- 4.2 Convolutional Pixel-Aligned Features -- 4.3 Multi-view Feature Fusion -- 4.4 Modeling Radiance Fields -- 5 Novel View Synthesis -- 5.1 Training and Implementation Details -- 6 Experiments -- 6.1 Reconstruction of Human Heads from Studio Data -- 6.2 Reconstruction from In-the-wild Captures -- 6.3 Reconstruction of Human Bodies -- 7 Conclusion -- References -- ViewFormer: NeRF-Free Neural Rendering from Few Images Using Transformers -- 1 Introduction -- 2 Related Work -- 3 Method -- 4 Experiments -- 5 Conclusions and Future Work -- References -- L-Tracing: Fast Light Visibility Estimation on Neural Surfaces by Sphere Tracing -- 1 Introduction. ; 2 Related Works -- 2.1 Neural Shape Representation -- 2.2 Reflectance and Illumination Estimation -- 3 Light Visibility Estimation -- 3.1 Preliminaries -- 3.2 L-Tracing -- 4 Reflectance Factorization Based on L-Tracing -- 4.1 Shape Learning -- 4.2 Reflectance Factorization -- 5 Experiments -- 5.1 Light Visibility Estimation -- 5.2 Reflectance Factorization -- 5.3 Novel View Interpolation and Relighting -- 5.4 Ablation Study -- 6 Conclusions -- References -- A Perceptual Quality Metric for Video Frame Interpolation -- 1 Introduction -- 2 Related Work -- 3 Video Frame Interpolation Quality Dataset -- 3.1 Data Preparation -- 3.2 Annotation -- 4 Video Frame Interpolation Quality Metric -- 5 Experiments -- 5.1 Comparisons to Existing Metrics -- 5.2 Ablation Studies -- 6 Limitations and Future Work -- 7 Conclusion -- References -- Adaptive Feature Interpolation for Low-Shot Image Generation -- 1 Introduction -- 2 Related Work -- 2.1 Low-Shot Data Generation -- 2.2 Geometric Interpretations in GANs -- 2.3 Interpolation in Feature Space and Mixup -- 3 Methodology -- 3.1 The Flattening Effect of Discriminator -- 3.2 Implicit Data Augmentation -- 3.3 Nearest Neighbors Interpolation -- 3.4 Data-Driven Adaptive Augmentation -- 4 Experiments -- 4.1 Datasets and Implementation Details -- 4.2 Ablation Study -- 4.3 Results -- 5 Discussion -- References -- PalGAN: Image Colorization with Palette Generative Adversarial Networks -- 1 Introduction -- 2 Related Work -- 2.1 Colorization -- 2.2 GAN-Based Image-to-Image Translation -- 3 Method -- 3.1 Palette Generator -- 3.2 Palette Assignment Generator -- 3.3 Color Discriminator -- 3.4 Learning Objective -- 4 Experiments -- 4.1 Implementation -- 4.2 Quantitative Evaluations -- 4.3 Qualitative Evaluations -- 4.4 Ablation Studies -- 5 Concluding Remarks -- References. ; Fast-Vid2Vid: Spatial-Temporal Compression for Video-to-Video Synthesis -- 1 Introduction -- 2 Related Work -- 3 Fast-Vid2Vid -- 3.1 A Revisit of GAN Compression -- 3.2 Overview of Fast-Vid2Vid -- 3.3 Spatial Resolution Compression for Vid2Vid -- 3.4 Temporal Sequential Data Compression for Vid2Vid -- 3.5 Semantic-Driven Motion Compensation for Interpolation Synthesis -- 4 Experiments -- 4.1 Experimental Setup -- 4.2 Quantitative Results -- 4.3 Ablation Study -- 4.4 Qualitative Results -- 5 Conclusion -- References -- Learning Prior Feature and Attention Enhanced Image Inpainting -- 1 Introduction -- 2 Related Work -- 3 Method -- 3.1 Masked Autoencoder for Inpainting -- 3.2 Attention-Based CNN Restoration (ACR) -- 3.3 Loss Functions -- 4 Experiments -- 5 Conclusions -- References -- Temporal-MPI: Enabling Multi-plane Images for Dynamic Scene Modelling via Temporal Basis Learning -- 1 Introduction -- 2 Related Work -- 3 Approach -- 3.1 The Multi-plane Image Representation -- 3.2 Temporal Basis Formulation -- 3.3 Temporal Coding for Novel-View Synthesis -- 3.4 Training Loss Function -- 4 Experiments -- 4.1 Implementation Details -- 4.2 Dataset -- 4.3 Ablation Study -- 4.4 Evaluation and Comparison -- 4.5 Baseline for Brute-Force Scenario -- 5 Concluding Remarks -- 5.1 Limitations -- 5.2 Conclusion -- References -- 3D-Aware Semantic-Guided Generative Model for Human Synthesis -- 1 Introduction -- 2 Related Work -- 3 Preliminaries -- 4 The Proposed 3D-SGAN -- 4.1 3D Generator for Semantic Mask Rendering -- 4.2 VAE-Conditioned Texture Generator -- 4.3 Consistency Losses for Semantics and Pose Disentanglement -- 4.4 Training and Inference -- 4.5 Real Image Editing Using GAN Inversion -- 5 Experiments -- 5.1 Comparisons with State-of-the-Art Methods -- 5.2 Ablation Study -- 5.3 Real Human Image Editing -- 6 Conclusion -- References. ; Temporally Consistent Semantic Video Editing.
• Titolo autorizzato: Computer Vision – ECCV 2022
• ISBN: 3-031-19784-4
• Formato: Materiale a stampa
• Livello bibliografico: Monografia
• Lingua di pubblicazione: Inglese
• Record Nr.: 996495565903316
• Lo trovi qui: Univ. di Salerno
• Serie: Lecture notes in computer science.