Computer Vision - ECCV 2022 Workshops

Name: Computer Vision - ECCV 2022 Workshops | Tel Aviv, Israel, October 23-27, 2022, Proceedings, Part II
Brand: Springer
Price: 171.19 EUR
Availability: OnlineOnly

Tel Aviv, Israel, October 23-27, 2022, Proceedings, Part II

Leonid Karlinsky Tomer Michaeli Ko Nishino(Herausgeber*in)

Springer (Verlag)

Erschienen am 15. Februar 2023

XXV, 765 Seiten

E-Book

PDF mit Wasserzeichen-DRM

Systemvoraussetzungen

978-3-031-25063-7 (ISBN)

171,19 €inkl. 7% MwSt.

Systemvoraussetzungen

für PDF mit Wasserzeichen-DRM

E-Book Einzellizenz

Als Download verfügbar

Beschreibung

The 8-volume set, comprising the LNCS books 13801 until 13809, constitutes the refereed proceedings of 38 out of the 60 workshops held at the 17th European Conference on Computer Vision, ECCV 2022. The conference took place in Tel Aviv, Israel, during October 23-27, 2022; the workshops were held hybrid or online.

The 367 full papers included in this volume set were carefully reviewed and selected for inclusion in the ECCV 2022 workshop proceedings. They were organized in individual parts as follows:

Part I: W01 - AI for Space; W02 - Vision for Art; W03 - Adversarial Robustness in the Real World; W04 - Autonomous Vehicle Vision

Part II: W05 - Learning With Limited and Imperfect Data; W06 - Advances in Image Manipulation;

Part III: W07 - Medical Computer Vision; W08 - Computer Vision for Metaverse; W09 - Self-Supervised Learning: What Is Next?;

Part IV: W10 - Self-Supervised Learning for Next-Generation Industry-LevelAutonomous Driving; W11 - ISIC Skin Image Analysis; W12 - Cross-Modal Human-Robot Interaction; W13 - Text in Everything; W14 - BioImage Computing; W15 - Visual Object-Oriented Learning Meets Interaction: Discovery, Representations, and Applications; W16 - AI for Creative Video Editing and Understanding; W17 - Visual Inductive Priors for Data-Efficient Deep Learning; W18 - Mobile Intelligent Photography and Imaging;

Part V: W19 - People Analysis: From Face, Body and Fashion to 3D Virtual Avatars; W20 - Safe Artificial Intelligence for Automated Driving; W21 - Real-World Surveillance: Applications and Challenges; W22 - Affective Behavior Analysis In-the-Wild;

Part VI : W23 - Visual Perception for Navigation in Human Environments: The JackRabbot Human Body Pose Dataset and Benchmark; W24 - Distributed Smart Cameras; W25 - Causality in Vision; W26 - In-Vehicle Sensing and Monitorization; W27 - Assistive Computer Vision and Robotics; W28 - Computational Aspectsof Deep Learning;

Part VII: W29 - Computer Vision for Civil and Infrastructure Engineering; W30 - AI-Enabled Medical Image Analysis: Digital Pathology and Radiology/COVID19; W31 - Compositional and Multimodal Perception;

Part VIII: W32 - Uncertainty Quantification for Computer Vision; W33 - Recovering 6D Object Pose; W34 - Drawings and Abstract Imagery: Representation and Analysis; W35 - Sign Language Understanding; W36 - A Challenge for Out-of-Distribution Generalization in Computer Vision; W37 - Vision With Biased or Scarce Data; W38 - Visual Object Tracking Challenge.

Weitere Details

Weitere Ausgaben

Inhalt

Intro
Foreword
Preface
Organization
Contents - Part II
W05 - Learning With Limited and Imperfect Data
W05 - Learning With Limited and Imperfect Data
SITTA: Single Image Texture Translation for Data Augmentation
1 Introduction
2 Related Work
3 Method
3.1 Workflow
3.2 Model Design
3.3 Loss Function
4 Experiments
4.1 Experimental Setup
4.2 Low-level Evaluation on Augmented Data
4.3 Augmented Data for Image Classification
5 Discussion
6 Conclusion
References
Learning from Noisy Labels with Coarse-to-Fine Sample Credibility Modeling
1 Introduction
2 Related Works
3 Method
3.1 Coarse Level Separation
3.2 Fine-grained Sequential Credibility Modeling
3.3 Selective Label Distribution Learning
4 Experiments
4.1 Datasets and Implementation Details
4.2 Comparison with State-of-the-Art Methods
4.3 Ablation Studies
5 Conclusion
References
PLMCL: Partial-Label Momentum Curriculum Learning for Multi-label Image Classification
1 Introduction
2 Related Work
2.1 Partial-label Learning Under Different Settings
2.2 Semi-supervised Learning
2.3 Curriculum Learning
3 The PLMCL Method
3.1 Overview
3.2 Updating Pseudo Labels by Momentum
3.3 Scheduled Loss Function
4 Experiments
4.1 Datasets Preparation
4.2 Subset Single Positive Label (SSPL)
4.3 Full-set Single Positive Label (FSPL)
4.4 Ablation Study
5 Conclusions
References
Open-Vocabulary Semantic Segmentation Using Test-Time Distillation
1 Introduction
2 Related Works
3 Method
3.1 Relevance Map Generation
3.2 From View-Averaged Relevance Maps to Segmentation
4 Experiments
4.1 Datasets and Evaluation Metrics
4.2 Baselines
4.3 Implementation Details
5 Results
5.1 Comparison to Other Methods
5.2 Analysis
6 Discussion
7 Conclusions
References
SW-VAE: Weakly Supervised Learn Disentangled Representation via Latent Factor Swapping
1 Introduction
2 Related Work
3 SW-VAE Model
4 Experimental Evaluation
4.1 Benchmarks, Baseline Methods and Evaluation Metrics
4.2 Quantitative Results
4.3 Ablation Study
5 Conclusion
References
Learning Multiple Probabilistic Degradation Generators for Unsupervised Real World Image Super Resolution
1 Introduction
2 Related Works
2.1 GAN-Based Unsupervised Real-World Super Resolution
2.2 Stochastic Degradation Operations
2.3 Noise Injection to Feature Space
3 Proposed Methods
3.1 Probabilistic Degradation Generator
3.2 Training Multiple Degradation Generators
3.3 Collaborative Learning
4 Experiments
4.1 Implementation Details
4.2 Ablation Studies
4.3 Comparison with State-of-the-Arts
4.4 Robustness to Perturbation
4.5 Discussion
5 Conclusion
References
Out-of-Distribution Detection Without Class Labels
1 Introduction
1.1 Related Works
2 Preliminaries
2.1 Deep Image Clustering
2.2 Feature-Adaptation for Labeled Multi-class Anomaly Detection
3 Deep Clustering for Multi-class Anomaly Detection
3.1 Self-Supervised Clustering for Multi-class Anomaly Detection
3.2 Finetuning Pre-trained Features with Pseudo Labels
4 Results
4.1 Multi-class Anomaly Detection with Pre-Trained Features
4.2 Multi-class Anomaly Detection Without Pretraining
4.3 Implementation Details
5 Discussion
6 Limitations
7 Conclusion
References
Unsupervised Domain Adaptive Object Detection with Class Label Shift Weighted Local Features
1 Introduction
2 Related Work
3 Preliminary: Marginal Feature Alignment
4 Method
4.1 Feature Map Decomposition
4.2 Class-aware Re-Weighting from a Mathematical View
4.3 Target Domain Class Label Distribution Estimation
4.4 Overview on the Proposed Iterative Approach
4.5 Proposed Modifications of Recent Methods
5 Experiments
5.1 Pascal VOC - Clipart
5.2 Sim10k - Cityscapes
5.3 Synscapes - Cityscapes
5.4 Ablation Study
6 Conclusion
References
OpenCoS: Contrastive Semi-supervised Learning for Handling Open-Set Unlabeled Data
1 Introduction
2 Preliminaries
2.1 Semi-supervised Learning
2.2 Contrastive Representation Learning
3 OpenCoS: A Framework for Open-Set SSL
3.1 Overview of OpenCoS
3.2 Detection Criterion
3.3 Auxiliary Loss and Batch Normalization
4 Experiments
4.1 Effects of Out-of-Class Unlabeled Samples
4.2 Experiments on CIFAR Datasets
4.3 Experiments on ImageNet Datasets
5 Ablation Study
6 Discussion
References
Semi-supervised Domain Adaptation by Similarity Based Pseudo-Label Injection
1 Introduction
2 Related Works
3 Method
3.1 Inter-domain Feature Alignment
3.2 Pseudo-Label Injection
3.3 Instance Level Similarity
3.4 Intra-domain Alignment
3.5 Classification Loss and Overall Framework
4 Experiments
4.1 Datasets
4.2 Baselines
4.3 Implementation Details
4.4 Results
5 Ablation Study
6 Conclusions
References
W06 - Advances in Image Manipulation
W06 - Advances in Image Manipulation
Evaluating Image Super-Resolution Performance on Mobile Devices: An Online Benchmark
1 Introduction
2 Related Work
2.1 Super-Resolution Methods
2.2 Mobile Super-Resolution Challenges and Benchmarks
3 Evaluating Image Super-Resolution on Mobile Devices
3.1 Overall Pipeline of Benchmark
3.2 Components of Benchmark System
3.3 Benchmark Protocols
4 Experiments
4.1 Implementation Details
4.2 Benchmark Results
4.3 Analysis of Architecture
5 Conclusion
References
Style Adaptive Semantic Image Editing with Transformers
1 Introduction
2 Related Work
3 Methodology
3.1 Problem Statement
3.2 Intra-/Inter-Image Knowledge Transfer
3.3 Generative Transformer Structure for SASIE
3.4 Transformer Decoder Layer
3.5 Training Loss and Objective
4 Experiments
4.1 Experimental Setup
4.2 Experimental Results
5 Conclusion
References
Third Time's the Charm? Image and Video Editing with StyleGAN3
1 Introduction
2 Related Work
3 The StyleGAN3 Architecture
4 Analysis
4.1 Rotation Control
4.2 Disentanglement Analysis
5 Image Editing
6 StyleGAN3 Inversion
6.1 Designing the Encoder Network
6.2 Inverting Images into StyleGAN3
7 Inverting and Editing Videos with StyleGAN3
8 Conclusions
References
CNSNet: A Cleanness-Navigated-Shadow Network for Shadow Removal
1 Introduction
2 Related Work
2.1 Shadow Removal
2.2 Region-Wise Information
2.3 Vision Transformer
3 Method
3.1 Cleanness-Navigated-Shadow Network
3.2 Soft-Region Mask Predictor
3.3 Shadow-Oriented Adaptive Normalization (SOAN)
3.4 Shadow-Aware Aggregation with Transformer (SAAT)
3.5 Loss Functions
4 Experiments
4.1 Datasets and Evaluation Measurements
4.2 Shadow Removal Evaluation on ISTD Dataset
4.3 Shadow Removal Evaluation on SRD Dataset
4.4 Ablation Studies
5 Conclusions
References
Unifying Conditional and Unconditional Semantic Image Synthesis with OCO-GAN
1 Introduction
2 Related Work
3 Method
3.1 Hybrid OCO-GAN Architecture
3.2 Training
4 Experiments
4.1 Experimental Setup
4.2 Main Results
4.3 Ablation Studies
5 Conclusion
References
Efficient Image Super-Resolution Using Vast-Receptive-Field Attention
1 Introduction
2 Related Work
3 Motivation
3.1 Large Kernel in Visual Attention
3.2 Parameter Reduction
3.3 Pixel Normalization for Stable Attention Training
3.4 Discussion
4 Network Architecture
4.1 The Building Blocks
5 Experiments
5.1 Experimental Setup
5.2 Comparison with State-of-the-Art Methods
5.3 Ablation Study on Micro Design
6 Conclusions
References
Unsupervised Scene Sketch to Photo Synthesis
1 Introduction
2 Related Work
3 Methods
3.1 Domain Standardization
3.2 Reference-Guided Photo Synthesis
4 Experimental Results
4.1 Network Architectures and Training Details
4.2 Datasets
4.3 Representation Encoding
4.4 Photo Synthesis
4.5 Human Perceptual Study
4.6 Photo Editing Through Sketch
4.7 Analysis and Ablation Studies
5 Summary
References
U-shape Transformer for Underwater Image Enhancement
1 Introduction
2 Related Work
2.1 Data-driven UIE Methods
2.2 Underwater Image Datasets
2.3 Transformers
3 Proposed Dataset and Method
3.1 LSUI Dataset
3.2 U-shape Transformer
3.3 Loss Function
4 Experiments
4.1 Experiment Settings
4.2 Dataset Evaluation
4.3 Network Architecture Evaluation
4.4 Ablation Study
5 Conclusions
References
Hybrid Transformer Based Feature Fusion for Self-Supervised Monocular Depth Estimation
1 Introduction
2 Related Works
3 Method
3.1 Depth Estimation Network
3.2 Pose Estimation Network
3.3 Training Strategies
4 Experiments
4.1 Dataset Details
4.2 Training and Implementation Details
4.3 Quantitative Analysis
4.4 Qualitative Analysis
4.5 Ablation Studies
5 Conclusion
References
Towards Real-World Video Deblurring by Exploring Blur Formation Process
1 Introduction
2 Related Works
2.1 Deep Deblurring
2.2 Motion Blur Datasets
3 Blur Formation Process Revisit
3.1 Real-World Blur Formation
3.2 Existing Blur Synthesis Pipeline
3.3 Synthetic Blur Analysis
4 RAW-Blur Dataset Construction
4.1 Ultra-High Frame-rate RAW Video Dataset
4.2 RAW-Blur Synthesis Pipeline
5 Experiments
5.1 Experimental Setting
5.2 Comparison to Existing Synthetic Datasets
5.3 Analysis of RAW-Blur Synthesis Pipeline
6 Conclusion and Limitation
References
Unified Transformer Network for Multi-Weather Image Restoration
1 Introduction
2 Related Work
3 Proposed Method
3.1 Original Resolution Transformer Stream
3.2 Multi-level Feature Aggregation Stream
3.3 Training Strategy
3.4 Loss Functions
4 Experimental Analysis
4.1 Datasets
4.2 Training Details
4.3 Quantitative and Qualitative Analysis
4.4 Ablation Study
5 Real World Application
6 Conclusion
References
DSR: Towards Drone Image Super-Resolution
1 Introduction
2 Related Work
3 DSR Dataset
3.1 Data Acquisition
3.2 Image Registration
3.3 Alignment Analysis
3.4 Data Analysis
4 Evaluating SR on Drone Data
4.1 Pretrained Off-the-Shelf SR Methods
4.2 Fine-Tuned SR Methods
4.3 Altitude-Aware SR
5 Conclusion and Future Directions
References
CEN-HDR: Computationally Efficient Neural Network for Real-Time High Dynamic Range Imaging
1 Introduction
2 Related Works
2.1 Deep Learning Based HDR Merging
2.2 Efficient Learning-Based HDR Merging Architectures
3 Proposed Method
3.1 Feature Encoding
3.2 Attention Module
3.3 Features Merging
3.4 Features Decoding
4 Experimental Settings
4.1 Datasets
4.2 Loss Function
4.3 Implementation Details
5 Experimental Results
5.1 Fidelity Performance
5.2 Efficiency Comparison
6 Conclusions
References
Image Super-Resolution with Deep Variational Autoencoders
1 Introduction
2 Related Work
3 Preliminaries
3.1 Variational Autoencoders
3.2 Conditional Variational Autoencoders
3.3 Very Deep Variational Autoencoder (VDVAE)
4 VDVAE-SR
4.1 LR-encoder
4.2 Conditional Prior
4.3 Generative Model and Inference Network
5 Experiments
5.1 Datasets
5.2 Implementation Details
5.3 Evaluation
5.4 Results
6 Conclusions
References
Light Field Angular Super-Resolution via Dense Correspondence Field Reconstruction
1 Introduction
2 Related Works
2.1 Non-Depth Based Methods
2.2 Depth Based Methods
3 Method
3.1 Light Field Representation
3.2 Proposed Network
3.3 Training Details
4 Experimental Results
4.1 Synthetic Scenes
4.2 Real-World Scenes
4.3 Visualization of Correspondence Maps
5 Conclusion
References
Adaptive Mask-Based Pyramid Network for Realistic Bokeh Rendering
1 Introduction
2 Related Work
3 Adaptive Mask-based Pyramid Network
3.1 Motivation
3.2 Overview
3.3 Mask-Guided Bokeh Generator Block
3.4 Laplacian Pyramid Refinement Block
3.5 Losses
4 Experiments
4.1 Dataset
4.2 Experimental Setup
4.3 Comparison with State-of-the-Art
4.4 Ablation Studies
5 Conclusion
References
RISPNet: A Network for Reversed Image Signal Processing
1 Introduction
2 Related Work
3 Method
3.1 RISPNet
3.2 RISPBlock
3.3 Third-Order Attention
3.4 Simple Activation Gate
4 Experiments
4.1 Implementation Details
4.2 AIM 2022 Reversed ISP Challenge Results
4.3 Ablation Study
4.4 Limitations
5 Conclusion
References
CIDBNet: A Consecutively-Interactive Dual-Branch Network for JPEG Compressed Image Super-Resolution
1 Introduction
2 Related Work
2.1 JPEG Compressed Image Super-resolution
2.2 Transformers in Low-Level Tasks
2.3 Introducing Convolutions to Transformers
3 Proposed Method
3.1 The Overall Architecture
3.2 Transformer Group
3.3 Convolution Group
3.4 Adaptive Cross-branch Fusion Module
4 Experiments
4.1 Datasets
4.2 Implementation Details
4.3 Ablation Study
4.4 Effects of ImageNet Training
4.5 Comparison with State-of-the-Art Methods
4.6 AIM Compressed Image Super-Resolution Challenge
5 Conclusions
References
XCAT - Lightweight Quantized Single Image Super-Resolution Using Heterogeneous Group Convolutions and Cross Concatenation
1 Introduction
2 Related Works
3 Method
3.1 XCAT's Architecture
3.2 Training and Quantization Details
4 Experimental Results
5 Conclusions and Future Studies
References
Learned Reverse ISP with Soft Supervision
1 Introduction
2 Related Work
2.1 Reversed ISP
2.2 Image Denoising
3 Proposed Method
3.1 Soft Supervision
3.2 SSDNet
4 Experimental Results
4.1 Experimental Setup
4.2 Ablation Studies
4.3 RAW Image Reconstruction
4.4 Application to Image Denoising
5 Conclusions
References
LiteDepth: Digging into Fast and Accurate Depth Estimation on Mobile Devices
1 Introduction
2 Related Work
3 Method
3.1 Network Design
3.2 R2 Crop
3.3 Multiple Loss Training
3.4 Structure-Aware Distillation
4 Experiments
4.1 Setup
4.2 Implementation Details
4.3 Quantitative Results
4.4 Qualitative Results
4.5 Inference Time
4.6 Ablation Studies
5 Conclusion
References
MSSNet: Multi-Scale-Stage Network for Single Image Deblurring
1 Introduction
2 Related Work
3 Shortcomings of Previous Coarse-to-Fine Approaches
4 Multi-Scale-Stage Network
4.1 Network Architecture
4.2 Training and Loss Functions
5 Experiments
5.1 Comparison with Previous Methods
5.2 Ablation Study and Analysis
6 Conclusion
References
RCBSR: Re-parameterization Convolution Block for Super-Resolution
1 Introduction
2 Related Work
2.1 Light-Weight Image Super-Resolution
2.2 Re-parameterization
2.3 Light-Weight Video Super-Resolution
3 Approach
3.1 Optimizing Network Architecture
3.2 NAS
3.3 Optimizing Training Strategy
4 Experiments
4.1 Dataset
4.2 Implementation Details
4.3 Comparisons with Bicubic Method
4.4 Results of the Real-Time VSR Challenge
5 Conclusion
References
Multi-patch Learning: Looking More Pixels in the Training Phase
1 Introduction
2 Related Work
2.1 Deep CNN for Image Super-Resolution
2.2 Vision Transformer
3 Methods
3.1 Flickr2K-L Dataset
3.2 Multi-patch Training Strategy
4 Experiments
4.1 Implementation Detail
4.2 Quantitative Results
4.3 Qualitative Results
4.4 Ablation Study
4.5 AIM 2022 Challenge
5 Conclusions
References
Fast Nearest Convolution for Real-Time Efficient Image Super-Resolution
1 Introduction
2 Related Work
2.1 Single Image Super-Resolution
2.2 Efficient Image Super-Resolution
3 Method
3.1 Network Architecture Selection
3.2 Nearest Convolution
3.3 Residual Learning
4 Experiments
4.1 Dataset and Implementation Details
4.2 Model Quantization
4.3 Experimental Results
5 Conclusion
References
Real-Time Channel Mixing Net for Mobile Image Super-Resolution
1 Introduction
2 Related Work
2.1 Lightweight Image Super-resolution
2.2 Mobile Image Super-Resolution
2.3 Network Quantization
3 Methodology
3.1 Channel and Deep Features Mixing (CDFM) Block
3.2 The CDFM-Mobile Framework
3.3 The Loss Function
4 Experiments
4.1 Settings
4.2 Comparison with State-of-the-Arts
4.3 MAI2022 SISR Challenge
4.4 Quantize-Aware Training
4.5 Ablation Study
5 Conclusion
References
Sliding Window Recurrent Network for Efficient Video Super-Resolution
1 Introduction
2 Related Works
2.1 Single Image Super-Resolution
2.2 Video Super-Resolution
2.3 Efficient Super-Resolution
3 Method
3.1 Sliding-Window Recurrent Network
3.2 Architecture and Loss
4 Experiment
4.1 Dataset
4.2 Implementation Details
4.3 Experimental Result
4.4 Ablation Study
5 Conclusion
References
EESRNet: A Network for Energy Efficient Super-Resolution
1 Introduction
2 Related Work
2.1 Single Image Super-Resolution
2.2 Multi-frame Super-Resolution
2.3 Mobile Super-Resolution
3 Method
3.1 EESRNet
3.2 TEESRNet
3.3 SEESRNet
4 Experiments
4.1 Settings
4.2 Comparison with the State-of-the-Art Methods
4.3 Ablation Study
4.4 Test Results
5 Conclusions
References
Bokeh-Loss GAN: Multi-stage Adversarial Training for Realistic Edge-Aware Bokeh
1 Introduction
2 Related Work
2.1 Computational Bokeh
2.2 Monocular Depth Estimation
2.3 Generative Models
3 Proposed Method
3.1 Backbone Network
3.2 Depth Estimation Network
3.3 Discriminator Loss
3.4 Bokeh Loss
3.5 Training Setup
4 Experiments
4.1 Experimental Setup
4.2 Quantitative and Qualitative Evaluation
4.3 Ablation Studies
5 Conclusions
A Pytorch to TFLite Conversion
References
Residual Feature Distillation Channel Spatial Attention Network for ISP on Smartphone
1 Introduction
2 Related Work
2.1 Learned Image Signal Processing
2.2 Re-parameterization
3 Proposed Method
3.1 Problem Formulation
3.2 Network Architecture
3.3 Loss Function
4 Experiment
4.1 Dataset
4.2 Evaluation Metrics
4.3 Implementation Details
4.4 Quantitative and Qualitative Evaluation
4.5 Ablation Study and AI Benchmark
4.6 Test Results of Mobile AI 2022 Learned Smartphone ISP Challenge
5 Conclusion
References
HST: Hierarchical Swin Transformer for Compressed Image Super-Resolution
1 Introduction
2 Related Works
2.1 Single Image Super-Resolution
2.2 Compression Artifacts Removal
3 Method
3.1 Hierarchical Feature Extraction
3.2 Feature Enhancement and Fusion
3.3 HR Reconstruction Module
3.4 Pretraining with SR
3.5 Loss Functions
4 Experiments
4.1 Datasets
4.2 Implementation Details
4.3 Effects of Different Pretraining Schemes
4.4 Effects of Hierarchical Architecture
4.5 Comparison with Other Frameworks
4.6 AIM2022 Challenge
5 Conclusion
References
Swin2SR: SwinV2 Transformer for Compressed Image Super-Resolution and Restoration
1 Introduction
2 Related Work
2.1 Image Restoration
2.2 Vision Transformer
3 Our Method
3.1 Experimental Setup
3.2 Implementation Details
4 Experimental Results
4.1 JPEG Compression Artifacts Removal
4.2 Classical Image Super-Resolution
4.3 Real-World Image Super-Resolution
4.4 Compressed Image Super-Resolution
5 Conclusion
References
Reversing Image Signal Processors by Reverse Style Transferring
1 Introduction
2 Related Works
2.1 Mapping from sRGB to RAW
2.2 Reverse Style Transferring
3 Proposed Methodology
4 Experiments and Results
4.1 Datasets
4.2 Experimental Details
4.3 Quantitative Results
4.4 Qualitative Comparison
5 Conclusion
References
Overexposure Mask Fusion: Generalizable Reverse ISP Multi-step Refinement
1 Introduction
2 Related Works
3 Methodology
3.1 Network Architecture
3.2 Reconstructing Demosaiced RAW
3.3 YUV Overexposure Mask
3.4 Overexposure Mask Fusion for Inference
3.5 Perceptual Loss Functions and Training Details
3.6 Bayer to Bayer Optional Refinement
4 Experiments
4.1 Quantitative and Qualitative Evaluations
4.2 Ablation Studies
4.3 Limitations
5 Conclusion
References
CAIR: Fast and Lightweight Multi-scale Color Attention Network for Instagram Filter Removal
1 Introduction
2 Backgrounds
2.1 Nonlinear Activation Free Network
2.2 Color Attention Mechanism
2.3 Ensemble Learning
3 Multi-scale Color Attention Network
3.1 Proposed CAIR Architecture
3.2 Color Attention Module
4 Experiments
4.1 Datasets
4.2 Implementation Details
4.3 Compared Methods
4.4 Ensemble Learning
4.5 Results
5 Conclusion
References
MicroISP: Processing 32MP Photos on Mobile Devices with Deep Learning
1 Introduction
2 Fujifilm UltraISP Dataset
3 Architecture
4 Experiments
4.1 Qualitative Evaluation
4.2 Quantitative Evaluation
4.3 Runtime Evaluation
4.4 Inference on Mobile NPUs
4.5 Ablation Study
4.6 Adjusting the Model Complexity
4.7 Limitations
4.8 PyNET-V2 Mobile
5 Conclusion
References
Real-Time Under-Display Cameras Image Restoration and HDR on Mobile Devices
1 Introduction
2 Related Work
3 Proposed Method
3.1 Method Description
4 Experimental Setup
4.1 Implementation Details
4.2 Experimental Results
5 Efficiency Analysis
6 Conclusions
References
Author Index

Systemvoraussetzungen

Als PDF speichern Als Link merken