MultiMedia Modeling : 29th International Conference, MMM 2023, Bergen, Norway, January 9–12, 2023, Proceedings, Part II

The two-volume set LNCS 13833 and LNCS 13834 constitutes the proceedings of the 29th International Conference on MultiMedia Modeling, MMM 2023, which took place in Bergen, Norway, during January 9-12, 2023. The 86 papers presented in these proceedings were carefully reviewed and selected from a total of 267 submissions. They focus on topics related to multimedia content analysis; multimedia signal processing and communications; and multimedia applications and services. Preface Organizing Committee Contents – Part II Contents – Part I Multimedia Processing and Applications Transparent Object Detection with Simulation Heatmap Guidance and Context Spatial Attention 1 Introduction 2 Related Work 3 Method 3.1 Simulation Heatmap Guidance 3.2 Context Spatial Attention 3.3 Network Architecture 4 Experiments 4.1 Main Results 4.2 Ablation Study 4.3 Visualization Analysis 5 Conclusion References Deep3DSketch+: Rapid 3D Modeling from Single Free-Hand Sketches 1 Introduction 2 Related Works 2.1 Sketch-Based 3D Modeling 2.2 Single-View 3D Reconstruction 3 Method 3.1 Preliminary 3.2 View-Aware and Structure-Aware 3D Modeling 3.3 Loss Function 4 Experiments 4.1 Datasets 4.2 Implementation Details 4.3 Results 4.4 Ablation Study 5 Conclusion References Manga Text Detection with Manga-Specific Data Augmentation and Its Applications on Emotion Analysis 1 Introduction 2 Manga Text Detection with Specific Data Augmentation 2.1 Overview 2.2 Network for Augmentation 2.3 Loss Functions for Augmentation 2.4 Augmented Manga Pages 3 Manga Emotion Analysis 4 Experiments of Manga Text Detection 4.1 Experimental Settings 4.2 Performance Evaluation 4.3 Ablation Studies 5 Experiments of Manga Emotion Recognition 6 Conclusion References SPEM: Self-adaptive Pooling Enhanced Attention Module for Image Recognition 1 Introduction 2 Related Works 3 Self-adaptive Pooling Enhanced Attention Module 3.1 Pooling Module 3.2 Excitation Module 3.3 Reweighting Module 3.4 Loss Function 4 Experiments 5 Ablation Study 6 Conclusion and Future Works References Less Is More: Similarity Models for Content-Based Video Retrieval 1 Introduction 2 Related Work 3 User Study 3.1 Dataset and Pre-processing 3.2 Similarity Models 3.3 Evaluation Procedure 4 Results 5 Discussion and Conclusions References Edge Assisted Asymmetric Convolution Network for MR Image Super-Resolution 1 Introduction 2 Related Work 2.1 Edge Assisted Image Super-Resolution 2.2 MR Image Super-Resolution 3 Methods 3.1 Network Architecture 3.2 Edge Assisted Feature Extraction Block 3.3 Asymmetric Convolution Group 3.4 Contextual Spatial Attention 4 Experiments 4.1 Datasets and Implementation Details 4.2 Model Analysis 4.3 Comparison with Other Methods 5 Conclusion References An Occlusion Model for Spectral Analysis of Light Field Signal 1 Introduction 2 Occlusion Framework 2.1 LF and Scene Surface Model 2.2 Problem Formulation 2.3 Determining the Occlusion Boundary Point 2.4 Determining of the Occlusion Function 3 Plenoptic Spectral Analysis 3.1 Specific Scene Surface and Light Ray Model 3.2 Light Field Bandwidth and Plenoptic Sampling Rate 4 Experimental Analysis 4.1 Public Dataset 4.2 LF Captured by Handheld Camera 5 Conclusion and Other Applications References Context-Guided Multi-view Stereo with Depth Back-Projection 1 Introduction 2 Related Works 2.1 Traditional Multi-view Stereo 2.2 Learning-Based Multi-view Stereo 2.3 Context Information 2.4 Back-Projection 3 Method 3.1 Overview of Coarse-to-Fine Multi-view Stereo 3.2 Context-Guided Cost Volume Regularization 3.3 Back-Projection Refined Up-Sampling (BPRU) 4 Experiment 4.1 Experiment Settings 4.2 Evaluation on DTU Dataset 4.3 Evaluation on Tanks and Temples Dataset 4.4 Ablation Studies 5 Conclusion References RLSCNet: A Residual Line-Shaped Convolutional Network for Vanishing Point Detection 1 Introduction 2 Related Work 3 Methods 3.1 Formulation 3.2 Residual Line-Shaped Convolutional Module 3.3 Network Implementation & Training Loss 4 Experimental Results and Analysis 4.1 Comparison with the State-of-the-Art 4.2 Discussions 5 Conclusion References Energy Transfer Contrast Network for Unsupervised Domain Adaption 1 Introduction 2 Related Work 2.1 Contrastive Learning 2.2 Energy Based Model 2.3 Unsupervised Domain Adaption 3 Proposed Method 3.1 Basic Definition 3.2 Contrastive Learning at the Instance Level 3.3 Energy Transfer 4 Experment 4.1 Datasets and Criteria 4.2 Implementation Details 4.3 Comparison with State of the Art 4.4 Ablation Studies 5 Conclusion References Recombining Vision Transformer Architecture for Fine-Grained Visual Categorization 1 Introduction 2 Related Work 2.1 Fine-Grained Visual Categorization 2.2 Vision Transformer 3 Method 3.1 The Overview of ViT 3.2 RecViT: A Recombination of Original ViT 3.3 Feature Processing Module 4 Experiment 4.1 Datasets and Implementation Details 4.2 Comparison on FGVC Benchmarks 4.3 Ablation Study 5 Conclusion References A Length-Sensitive Language-Bound Recognition Network for Multilingual Text Recognition 1 Introduction 2 Related Work 2.1 Scene Text Recognition 2.2 Multilingual Scene Text Recognition 3 Methodology 3.1 Language-Balanced Data Augmentation (LDA) 3.2 Script Identification Module (SIM) 3.3 Length-Sensitive Encoder (LE) 3.4 Language-Bound Decoder (LD) 3.5 Loss 4 Experiments 4.1 Datasets 4.2 Data Preprocessing 4.3 Comparisons with Other Methods 4.4 Ablation Experiments 5 Conclusion References Lightweight Multi-level Information Fusion Network for Facial Expression Recognition 1 Introduction 2 Related Work 2.1 FER Models 2.2 Lightweight Models 2.3 Lightweight FER Models 3 Method 3.1 Multi-level Information Fusion 3.2 Distillation Loss Function 4 Experiments 4.1 Datasets and Settings 4.2 Ablation Studies 4.3 Comparison with State-of-the-Art FER Methods 4.4 Visualization 5 Conclusion References Practical Analyses of How Common Social Media Platforms and Photo Storage Services Handle Uploaded Images 1 Introduction 2 Background and Related Work 2.1 Background 2.2 Related Work 3 Methods 4 Results 4.1 Data and Experimental Setup 4.2 Results on Google Photos 4.3 Results on Facebook 4.4 Flickr 5 Potential Applications 6 Conclusion References CCF-Net: A Cascade Center-Based Framework Towards Efficient Human Parts Detection 1 Introduction 2 Related Works 2.1 Lightweight Object Detection Methods 2.2 Distillation in Detector 3 Methodology 3.1 Adaptive Gaussian Mask 3.2 Cascade Attention Module 3.3 Center-Based Knowledge Distillation 3.4 Class-Specific Branch Mask 4 Experimental Results 4.1 Datasets 4.2 Implementation Details 4.3 Ablation Study 4.4 Comparisons with Lightweight Models 4.5 Comparisons with the State-of-Arts 5 Conclusion References Low-Light Image Enhancement Under Non-uniform Dark 1 Introduction 2 Related Work 2.1 Low-Light Image Enhancement 2.2 Low-Light Image Enhancement Dataset: The Status Quo 3 UDL Dataset 3.1 Dataset Creation 3.2 Data Comparison 4 Our Method 4.1 Non-uniform Dark Visual Network 4.2 Loss Functions 5 Experiment 5.1 Experimental Evaluation 5.2 Ablation Experiments 6 Conclusion References A Proposal-Improved Few-Shot Embedding Model with Contrastive Learning 1 Introduction 2 Related Work 2.1 Few-Shot Learning 2.2 Contrastive Learning 2.3 Object Detection 3 Method 3.1 Proposal Boxes 3.2 Enhanced Contrastive Learning 3.3 Training Phase 3.4 Inference Phase 4 Experimental Analysis 4.1 Datasets 4.2 Implementation Details 4.3 Ablation Experiment 4.4 Comparison with State-of-the-Art 5 Conclusion References Weighted Multi-view Clustering Based on Internal Evaluation 1 Introduction 2 Proposed Algorithm 2.1 Internal Evaluation Criterion 2.2 Determining 2.3 Weight Initialization 2.4 Weight Updating 3 Experiments 3.1 Datasets 3.2 Experimental Results and Comparison 4 Conclusion References BENet: Boundary Enhance Network for Salient Object Detection 1 Introduction 2 Related Work 3 Proposed Method 3.1 Progressive Feature Extraction Module 3.2 Adaptive Edge Fusion Module 3.3 Loss 4 Experiment 4.1 Dataset and Evaluation Metrics 4.2 Comparison 4.3 Ablation Studies 5 Conclusion References PEFNet: Positional Embedding Feature for Polyp Segmentation 1 Introduction 2 Related Work 3 Proposed Method 3.1 Architecture 3.2 Positional Embedding Feature Block 4 Experiment 4.1 Dataset 4.2 Implementation Details 4.3 Result 4.4 Ablation Study 5 Conclusion References MCOM-Live: A Multi-Codec Optimization Model at the Edge for Live Streaming 1 Introduction 2 Related Work 3 Multi-Codec Optimization Model 3.1 Motivation 3.2 Optimization Model Formulation 4 Experiments and Discussion 4.1 Experimental Setup 4.2 Experimental Results 5 Conclusions and Future Work References LAE-Net: Light and Efficient Network for Compressed Video Action Recognition 1 Introduction 2 Approach 2.1 Motion Temporal Fusion Module 2.2 Double Compress Knowledge Distillation 3 Experience 3.1 Datasets and Evaluation 3.2 Implementation Details 3.3 Ablation Studies 3.4 Comparison with the State of the Art 4 Conclusion References DARTS-PAP: Differentiable Neural Architecture Search by Polarization of Instance Complexity Weighted Architecture Parameters 1 Introduction 2 Method 2.1 Performance Collapse in DARTS 2.2 Introduction of Polarization Regularizer 2.3 Learning of Instance Complexity Weighted Architecture Parameters 3 Experiments 3.1 Training Setting 3.2 Searching Architectures for CIFAR-10 3.3 Results on Cifar-10 3.4 Transferring to ImageNet 3.5 Results on NAS-Bench-201 Search Space 4 Ablation Study 5 Conclusion References Pseudo-label Diversity Exploitation for Few-Shot Object Detection 1 Introduction 2 Related Works 3 Proposed Method 3.1 Problem Definition 3.2 Revitalized Pseudo-Margin Evaluation Loss 3.3 Novel Instance Bank 3.4 Correlation-Guided Loss Correction 4 Experiment 4.1 Benchmarks and Setups 4.2 Few-Shot Detection Results 4.3 Ablation Study 5 Conclusion References HSS: A Hierarchical Semantic Similarity Hard Negative Sampling Method for Dense Retrievers 1 Introduction 2 Background 3 Hierarchical Semantic Similarity (HSS) 3.1 Overview 3.2 Topic-Level: LDA-Based Sampling 3.3 Class-Level: SCCL-Based Sampling 3.4 Training 4 Experimental Setup 4.1 Datasets 4.2 Evaluation 4.3 Baselines 5 Experiments 5.1 Main Results 5.2 Ablation Study 5.3 QA Accuracy 6 Related Work 7 Conclusion and Future Work References Realtime Sitting Posture Recognition on Embedded Device 1 Introduction 2 Related Work 2.1 Sitting Posture Recognition 2.2 Model Pruning Methods 3 Methodology 3.1 Overall Structure Design 3.2 Optimization OpenPose 3.3 Design Sitting Posture Recognizer 4 Experiments 4.1 Experiment Environment 4.2 Constructing Datasets 4.3 Feature Fusion Results and Analysis 4.4 Pruning and Model Deployment 4.5 Comparison with Others 5 Conclusion References Comparison of Deep Learning Techniques for Video-Based Automatic Recognition of Greek Folk Dances 1 Introduction 2 Related Work 3 Dataset Description 4 Experiments and Results 4.1 Network Architectures and Input Representations 4.2 Training Procedure and Experimental Results 5 Conclusions References Dynamic Feature Selection for Structural Image Content Recognition 1 Introduction 2 Related Work 2.1 Multi-stage Approaches 2.2 End-to-End Approaches 3 Model 3.1 Model Overview 3.2 Fine-Grained Image Encoder 3.3 Dynamic Feature Selector 3.4 Spatial Relation Extractor 3.5 Transcribing Decoder 4 Experiment 4.1 Experimental Setup 4.2 Performance Comparison (RQ1) 4.3 Analysis on Dynamic Feature Selector (RQ2) 4.4 Ablation Study (RQ3) 4.5 Case Study 5 Conclusion References Dynamic-Static Cross Attentional Feature Fusion Method for Speech Emotion Recognition 1 Introduction 2 Methodology 2.1 Acoustic Feature Extraction 2.2 Dynamic-Static Cross Attentional Feature Fusion Method 2.3 Multi-label Auxiliary Learning 3 Experiment 3.1 Datasets 3.2 Experiment Setup 3.3 Experimental Results 3.4 Ablation Study 4 Conclusion References Research on Multi-task Semantic Segmentation Based on Attention and Feature Fusion Method 1 Introduction 2 Related Work 3 Network Model 3.1 Network Architecture 3.2 Attention Mechanism Module 3.3 Loss Functions 4 Experimental Results Section 4.1 Datasets 4.2 Baselines 4.3 Results and Analysis 5 Conclusions References Space-Time Video Super-Resolution 3D Transformer 1 Introduction 2 Related Work 3 Proposed Method 3.1 3D Pyramid Align Network 3.2 3D Transformer Network 3.3 Reconstruction Network 4 Loss Function 5 Experiments 5.1 Training Datasets 5.2 Training Details 5.3 Comparison to State-of-the-Art Methods 5.4 Ablation Studies 6 Conclusion References Graph-Based Data Association in Multiple Object Tracking: A Survey 1 Introduction 2 Related Work 3 Theoretical Background 4 Analysis of MOT Methods 4.1 Measurement-to-Measurement Association 4.2 Measurement-to-Track Association 4.3 Track-to-Track Association 5 Qualitative Comparison and Discussion 6 Conclusion References Multi-view Adaptive Bone Activation from Chest X-Ray with Conditional Adversarial Nets 1 Introduction 2 Related Work 3 Preliminary 3.1 DES Images 3.2 Dataset Preparation 4 Proposed Method 4.1 Overview 4.2 Supervisor 4.3 Activator 5 Experiments 5.1 Implementation Details 5.2 Metrics 5.3 Quantitative Analysis 5.4 Qualitative Evaluation 6 Conclusion References Multimodal Reconstruct and Align Net for Missing Modality Problem in Sentiment Analysis 1 Introduction 2 Related Work 2.1 Multimodal Sentiment Analysis 2.2 Missing Modality Problem 3 Proposed Method 3.1 Problem Definition 3.2 Feature Extraction Module 3.3 Feature Reconstruction Module 3.4 Text-centered Multimodal Alignment Module 3.5 Classification Module 3.6 Model Training 4 Experiments 4.1 Datsets 4.2 Data Preprocess 4.3 Implementation Details 4.4 Experimental Results 4.5 Ablation Study 5 Conclusion References Lightweight Image Hashing Based on Knowledge Distillation and Optimal Transport for Face Retrieval 1 Introduction 2 Proposed Lightweight Hashing 2.1 Attention-Based Triplet Knowledge Distillation 2.2 Hash Quantization Based on Optimal Transport 2.3 Alternate Training Strategy 3 Experiments 3.1 Experimental Setting 3.2 Performance Comparisons 3.3 Performance of Network Compression 4 Conclusions References CMFG: Cross-Model Fine-Grained Feature Interaction for Text-Video Retrieval 1 Introduction 2 Method 2.1 Visual Representation 2.2 Text Representation 2.3 Cross-Retrieval Module 2.4 Loss Function 3 Experiments 3.1 Datasets 3.2 Experimental Settings 3.3 Experimental Results 3.4 Ablation Study 4 Conclusion References Transferable Adversarial Attack on 3D Object Tracking in Point Cloud 1 Introduction 2 Related Work 3 The Proposed Methodology 3.1 3D Encoder-Decoder Adversarial Generator 3.2 Multi-fold Drift Loss 3.3 Overall Loss 4 Experiments 4.1 Overall Attack Performance 4.2 Ablation Study 4.3 Transfer to Unseen Trackers 5 Conclusion References A Spectrum Dependent Depth Layered Model for Optimization Rendering Quality of Light Field 1 Introduction 2 Spectrum Aliasing Analysis of Light Field 2.1 The Representation of the Light Field 2.2 Light Field Spectrum 2.3 Light Field Spectrum Aliasing 3 Spectrum Aliasing Eliminated by Deep Stratification 4 Experimental Results 4.1 Spectrum Simulation and Analysis 4.2 Comparison Reconstruction Results for Actual Scenes 5 Conclusion References Transformer-Based Cross-Modal Recipe Embeddings with Large Batch Training 1 Introduction 2 Related Work 2.1 Cross-Modal Recipe Retrieval 2.2 Food Image Synthesis 3 Method 3.1 Overview 3.2 Recipe Encoder and Self-supervised Learning 3.3 Image Encoder 3.4 Modality Alignment Loss 3.5 Retrieval Loss 3.6 Translation Consistency Loss 4 Experiments 4.1 Implementation Details 4.2 Cross-Modal Recipe Retrieval 4.3 Image Generation 4.4 Ablation Studies 4.5 Batch Size in Cross-Modal Embedding Learning 5 Conclusions References Self-supervised Multi-object Tracking with Cycle-Consistency 1 Introduction 2 Related Work 2.1 Self-supervised Learning over Videos 2.2 Multi-object Tracking 3 Approach 3.1 Overview 3.2 Tracker Model 3.3 Self-supervision 4 Experiments 4.1 Experimental Setup 4.2 Ablation Studies 4.3 MOTChallenge Results 4.4 Qualitative Results 5 Conclusion References Video-Based Precipitation Intensity Recognition Using Dual-Dimension and Dual-Scale Spatiotemporal Convolutional Neural Network 1 Introduction 2 Network Architecture 2.1 Dual-Dimension and Dual-Scale Spatiotemporal Convolutional Neural Network 2.2 Global Spatiotemporal Module, GSM 2.3 Local Spatiotemporal Module (LSM) 3 Experiments 3.1 Video-Based Precipitation Dataset 3.2 Experimental Environment and Evaluation Metric 3.3 Experimental Analysis 3.4 Ablation Experiment 4 Conclusions References Low-Light Image Enhancement Based on U-Net and Haar Wavelet Pooling 1 Introduction 2 Related Work 3 Methodology 3.1 Network Architecture 3.2 Loss Functions 4 Experimental Results 4.1 Datasets 4.2 Training Process 4.3 Results 5 Conclusions References Audio-Visual Sensor Fusion Framework Using Person Attributes Robust to Missing Visual Modality for Person Recognition 1 Introduction 2 Literature Review 3 Proposed Framework 3.1 Speech-to-Attribute Model 3.2 CTNet 4 Experiments 4.1 Experimental Setup 4.2 Baseline and Ablation Frameworks 4.3 Discussion 5 Conclusion References Rumor Detection on Social Media by Using Global-Local Relations Encoding Network 1 Introduction 2 Related Work 3 Preliminaries 3.1 Problem Definition 3.2 Decomposed Heterogeneous Graph 4 Methodology 4.1 Global Subgraph Encoding 4.2 Local Subgraph Encoding 4.3 Root Feature Enhancement 4.4 Rumor Detection 5 Experiments 5.1 Dataset and Metrics 5.2 Baselines 5.3 Comparison with Baselines 5.4 Ablation Experiments 5.5 Early Detection 6 Conclusions References Unsupervised Encoder-Decoder Model for Anomaly Prediction Task 1 Introduction 2 Related Work 3 Unsupervised Memory-Based Transformer for Anomaly Detection 3.1 Transformer Encoder 3.2 Memory Module 3.3 Transformer Decoder 3.4 Loss 3.5 Abnormality Score 4 Experimantal Results 4.1 Experimantal Setup 5 Results 5.1 Visualization Results 5.2 Visual Analysis of Different Scenes 5.3 Ablation Study 6 Conclusion References CTDA: Contrastive Temporal Domain Adaptation for Action Segmentation*-4pt 1 Introduction 2 Related Work 3 Methodology 3.1 Framework Overview 3.2 Self-supervised Learning Module 3.3 Feature Extraction Module 3.4 Implementation Details 4 Experiments 4.1 Datasets and Evaluation Metrics 4.2 Experimental Results 4.3 Ablation Study 4.4 Quantitative Effects 5 Conclusions References Multi-scale and Multi-stage Deraining Network with Fourier Space Loss 1 Introduction 2 Related Work 2.1 Single Image Deraining 2.2 Multi-scale Learning 2.3 Multil-stage Learning 3 Proposed Method 3.1 The Overall Structure of MS2DNet 3.2 Multi-scale Rain Streaks Extract Module 3.3 Background Recover Module 3.4 Training Loss for Deraining Network 4 Experiment and Result 4.1 DataSet 4.2 Implementation Details 4.3 Comparisons with the State-of-the-arts 4.4 Ablation Study 5 Conclusion References DHP: A Joint Video Download and Dynamic Bitrate Adaptation Algorithm for Short Video Streaming 1 Introduction 2 Related Work 3 DHP: A Joint Video Download and Dynamic ABR Algorithm 3.1 The Video Download Mechanism 3.2 Dynamic Bitrate Adaptation Algorithm 4 Performance Evaluation 4.1 Dataset and Evaluation Metrics 4.2 Result Analysis 5 Summary References Generating New Paintings by Semantic Guidance 1 Introduction 2 Related Work 2.1 Stroke-Based Rendering 2.2 Semantic Segmentation and Generation by Semantic Guidance 3 Approach 3.1 Preliminary 3.2 Overview 3.3 Painting with Semantic Guidance 3.4 Joint Optimization 4 Experiments 4.1 Experimental Settings 4.2 Results 4.3 Analysis 5 Conclusion References A Multi-Stream Fusion Network for Image Splicing Localization 1 Introduction 2 Related Work 3 Approach Overview 3.1 Multi-stream Architecture 3.2 Handcrafted Signals 3.3 Training Process 4 Experimental Setup 4.1 Datasets 4.2 Implementation Details 4.3 Evaluation Metrics 5 Experiments and Results 5.1 Ablation Study 5.2 Comparison with the State-of-the-Art 5.3 Qualitative Results 6 Conclusion References Fusion of Multiple Classifiers Using Self Supervised Learning for Satellite Image Change Detection 1 Introduction 2 Related Work 3 Methodology 3.1 Network Architecture 3.2 Datasets 3.3 Settings 4 Results 5 Conclusion References Improving the Robustness to Variations of Objects and Instructions with a Neuro-Symbolic Approach for Interactive Instruction Following 1 Introduction 2 Neuro-Symbolic Instruction Follower 2.1 Notation 2.2 Language Encoder 2.3 Visual Encoder 2.4 Semantic Understanding 2.5 MaskRCNN 2.6 Subtask Updater 2.7 Action Decoder 2.8 Object Selector 2.9 Progress Monitor 3 Experiments 3.1 Dataset 3.2 Training Details 3.3 Main Results 3.4 Performance of Semantic Understanding 4 Analysis: Evaluating the Robustness to Variations of Language Instructions 5 Related Work 5.1 Neuro-Symbolic Method 5.2 Embodied Vision-and-Language Task 6 Conclusion References Interpretable Driver Fatigue Estimation Based on Hierarchical Symptom Representations 1 Introduction 2 Related Works 3 Proposed Approach 3.1 The Proposed Framework 3.2 Hierarchical Fatigue Symptom Classification 3.3 Interpretable Driver Fatigue Estimation 4 Experiments 4.1 Dataset and Implementation Protocols 4.2 Comparison of Fatigue State Estimation 4.3 Ablation Study 5 Conclusion References VAISL: Visual-Aware Identification of Semantic Locations in Lifelog 1 Introduction 2 Related Work 2.1 GPS Trajectories Segmentation 2.2 Text-Image Embedding Models 3 Lifelog Dataset 4 The Proposed Method 4.1 Data Cleaning 4.2 Stop/Trip Point Classification 4.3 Post-processing 4.4 Event Characterisation 5 Experiments and Results 5.1 Event Detection Results 5.2 Event Characterisation Results 6 Discussions and Conclusion References Multi-scale Gaussian Difference Preprocessing and Dual Stream CNN-Transformer Hybrid Network for Skin Lesion Segmentation 1 Introduction 2 Proposed Approach 2.1 Network Architecture 2.2 Multi-scale Gaussian Difference 2.3 Dual Steam DNN Feature Extraction 2.4 Loss Function 3 Experiment 3.1 Dataset 3.2 Implementation Detail 3.3 Metrics 3.4 Quantitative Results on ISIC2016 and ISIC 2018 3.5 Qualitative Results on ISIC 2016 and 2018 3.6 Ablation Study 4 Conclusions References AutoRF: Auto Learning Receptive Fields with Spatial Pooling 1 Introduction 2 Related Work 2.1 Receptive Field Design 2.2 Attention Mechanism 2.3 Neural Architecture Search 3 Auto Learning Receptive Fields 3.1 Adaptive Receptive Fields Attention Module 3.2 Receptive Fields Search Space 4 Experiments 4.1 Experiment Setup 4.2 Image Classification 5 Conclusion References In-Air Handwritten Chinese Text Recognition with Attention Convolutional Recurrent Network 1 Introduction 2 Preprocessing 2.1 Removal of Redundant Points and Smoothing 2.2 Coordinate Transformation 2.3 Eigenvectors of Coordinates 3 Proposed Method 3.1 1DCNN 3.2 BLSTM Combined with Multihead Attention Mechanism 3.3 CTC Decoding 4 Experiments 4.1 Datasets 4.2 Model Training Environment 4.3 Evaluation Criteria 4.4 Experiment Results 5 Conclusion References BNI: Brave New Ideas Multimedia Datasets: Challenges and Future Possibilities 1 Introduction 2 Data Storage and Sharing 3 Modeling 3.1 Incorporating Information from Already Available Datasets 3.2 Incorporating Prior Knowledge into Data Modeling 4 Data Privacy 5 Conclusion and Future Research Directions References The Importance of Image Interpretation: Patterns of Semantic Misclassification in Real-World Adversarial Images 1 Introduction 2 Experiments on Transfer Adversarial Attacks 3 Evaluation Based on Semantic Mismatch 4 Results 5 Conclusion and Outlook References Research2Biz Students Take Charge of Climate Communication 1 Introduction 2 Background and Theories 3 Materials and Methods 4 Students Take Charge of Climate Communication 5 The Suitability of Innovation Pedagogy 6 Conclusion References Demo Social Relation Graph Generation on Untrimmed Video 1 Introduction 2 System Architecture 2.1 Video Process Engine 2.2 Relation Recognition Module 2.3 Graph Generation Module 3 Experiment 4 Conclusions and Future Work References Improving Parent-Child Co-play in a Roblox Game 1 Introduction 2 Related Works 3 Designing Mini-Games for Improved Co-play 4 User Study 4.1 User Study Procedure 4.2 User Study Results 5 Conclusions and Future Work References Taylor – Impersonation of AI for Audiovisual Content Documentation and Search 1 Introduction 2 System Overview 3 User Interfaces 3.1 Taylor 3.2 Annotation User Interface 3.3 Search User Interface 4 Conclusion References Virtual Try-On Considering Temporal Consistency for Videoconferencing 1 Introduction 2 Related Work 2.1 Virtual Try-On 3 Proposed Method 3.1 Virtual Fitting System Using a Webcam 3.2 Learning Considering Temporal Consistency 4 Experiments 4.1 Implementation Details 4.2 Experimental Results 5 Conclusion References Author Index