Pattern Recognition. ICPR International Workshops and Challenges

Name: Pattern Recognition. ICPR International Workshops and Challenges | Virtual Event, January 10-15, 2021, Proceedings, Part VI
Brand: Springer
Price: 96.29 EUR
Availability: OnlineOnly

Virtual Event, January 10-15, 2021, Proceedings, Part VI

Alberto Del Bimbo Rita Cucchiara Stan Sclaroff Giovanni Maria Farinella Tao Mei Marco Bertini Hugo Jair Escalante Roberto Vezzani(Herausgeber*in)

Springer (Verlag)

Erschienen am 24. Februar 2021

XX, 821 Seiten

E-Book

PDF mit Wasserzeichen-DRM

Systemvoraussetzungen

978-3-030-68780-9 (ISBN)

96,29 €inkl. 7% MwSt.

Systemvoraussetzungen

für PDF mit Wasserzeichen-DRM

E-Book Einzellizenz

Als Download verfügbar

Beschreibung

Weitere Details

Weitere Ausgaben

Inhalt

Intro
Foreword by General Chairs
Preface
Challenges
ICPR Organization
Contents - Part VI
MAES - Machine Learning Advances Environmental Science
Workshop on Machine Learning Advances Environmental Science (MAES)
Organization
MAES Chairs
Publicity Chair
Program Committee
Additional Reviewers
Finding Relevant Flood Images on Twitter Using Content-Based Filters
1 Introduction
2 Datasets and Search Objectives
2.1 The European Flood 2013 Dataset
2.2 Real-World Twitter Data
3 Methods
4 Experiments
4.1 Ranking Images by Relevance
4.2 On-Line Filter with Hard Decisions
5 Conclusions
References
Natural Disaster Classification Using Aerial Photography Explainable for Typhoon Damaged Feature
1 Introduction
1.1 Typhoon Damage Prediction for Immediate Response and Recovery
1.2 Related Works and Papers
1.3 Feature Extraction for Natural Disaster Damage Assessment
2 Modelling
2.1 Partition Clips and Learning Disaster Features
2.2 Visual Explanation Toward Disaster Features
3 Applied Results
3.1 Training and Test Dataset of Aerial-Photographs
3.2 Damage Feature Classifier Trained Results
3.3 Damage Feature Map and Unit Grid Visualization Results
4 Concluding Remarks
4.1 Disaster Features Visualization for Immediate Response Support
4.2 Future Works for Disaster Visual Mining and Learning Variations
Appendix
References
Environmental Time Series Prediction with Missing Data by Machine Learning and Dynamics Recostruction
1 Introduction
2 Model Order Estimation by Grassberger-Procaccia Algorithm
3 Support Vector Machine for Regression
4 Iterated Prediction and Imputation Algorithm
5 Experimental Results
6 Conclusions
References
Semi-Supervised Learning for Grain Size Distribution Interpolation
1 Introduction
2 Related Work
3 Research Area and Dataset
3.1 Target Variable: Grain Size Distribution
3.2 Auxiliary Data
4 Methodology
5 Experiments
5.1 Methods
5.2 Evaluation
6 Results
6.1 Analysis
7 Discussion
8 Conclusion
References
Location-Specific vs Location-Agnostic Machine Learning Metamodels for Predicting Pasture Nitrogen Response Rate
1 Introduction
2 Materials and Methods
2.1 Case Study, Data Description
2.2 Data Preprocessing
2.3 Machine Learning Pipeline
2.4 Evaluation
2.5 Implementation
3 Results
4 Discussion
5 Limitations
6 Conclusion and Future Work
References
Pattern Classification from Multi-beam Acoustic Data Acquired in Kongsfjorden
1 Introduction
2 Background
3 Materials and Methods
3.1 Clustering Method
4 Results
5 Discussion
6 Conclusions
References
Unsupervised Classification of Acoustic Echoes from Two Krill Species in the Southern Ocean (Ross Sea)
1 Introduction
2 Materials and Methods
2.1 Acoustic Data: Acquisition and Processing
2.2 Exploratory Analysis and Data Preparation
2.3 Clustering
3 Results
4 Discussion
5 Conclusions
References
Multi-Input ConvLSTM for Flood Extent Prediction
1 Introduction
2 Related Work
3 Method
4 Experiments
4.1 Homogeneous Data
4.2 Heterogeneous Data
5 Conclusion
References
Developing a Segmentation Modelpg for Microscopic Images of Microplastics Isolated from Clams
1 Introduction
2 Background
2.1 Measurement of Microplastics in Seafood Using MP-VAT
2.2 Image Segmentation and Deep Learning
3 Dataset Acquisition
3.1 Wet-Lab Phase
3.2 Dry-Lab Phase
4 Methods
4.1 Problem Definition
4.2 Dataset Characteristics
4.3 Model Training
4.4 Performance Metrics
5 Results and Discussion
6 Conclusions and Future Work
References
A Machine Learning Approach to Chlorophyll a Time Series Analysis in the Mediterranean Sea
1 Introduction
2 Materials and Methods
2.1 Data Collection and Pre-processing
2.2 SOM as a Missing Data Reconstruction Technique
2.3 SOM as a Classifier for Pattern Extraction
3 Results and Discussion
3.1 Missing Data Reconstruction
3.2 Classification of Chlorophyll a Time Series
4 Conclusions
References
Plankton Recognition in Images with Varying Size
1 Introduction
2 CNNs with Varying Image Size
3 Experiments
3.1 Data
3.2 CNN Architectures and Implementation Details
3.3 Spatial Pyramid Pooling
3.4 Metadata
3.5 Patch Cropping
3.6 Multi-stream CNN
3.7 Comparison of the Approaches
4 Conclusions
References
Environment Object Detection for Marine ARGO Drone by Deep Learning
1 Introduction
2 ARGO Drone
3 Object Detection Models
4 Experimental Results
5 Conclusions
References
Unsupervised Learning of High Dimensional Environmental Data Using Local Fractality Concept
1 Introduction
2 Method
3 Cases Study
4 Discussion and Conclusions
References
Spatiotemporal Air Quality Inference of Low-Cost Sensor Data
Application on a Cycling Monitoring Network
1 Introduction
1.1 Urban Air Quality
2 Material and Methods
2.1 Snuffelfiets
2.2 Data Inference in Space and Time
2.3 Model Performance
3 Results
4 Conclusions
References
How Do Deep Convolutional SDM Trained on Satellite Images Unravel Vegetation Ecology?
1 Introduction
2 Materials and Methods
2.1 CNN-SDM Model Training and Validation
2.2 Ecological Interpretation of the Learned Features
2.3 Environmental and Trait Data
2.4 Dimension Reduction
2.5 Ecological Interpretation of T-SNE Dimensions
2.6 Visualization
3 Results and Discussion
4 Conclusion
References
ManifLearn - Manifold Learning in Machine Learning, from Euclid to Riemann
Preface
Organization
General Chairs
Program Committee
Latent Space Geometric Statistics
1 Introduction
2 Latent Space Geometry
2.1 Latent Data Representations
2.2 Geodesics and Brownian Motions
3 Computational Representation
4 Nonlinear Latent Space Statistics
4.1 Fréchet and ML Means
4.2 Principal Component Analysis
4.3 Generalised Two-Sample Test
5 Maximum Likelihood Inference of Diffusions
5.1 Bridge Simulation and Parameter Inference
6 Experiments
6.1 MNIST
6.2 Diatoms
7 Conclusion
References
Improving Neural Network Robustness Through Neighborhood Preserving Layers
1 Introduction
2 Model Setup and Background
2.1 Model Setup
2.2 Adversarial Attack
2.3 Dimension Reduction with Neighborhood Preservation
3 A Novel Neighborhood Preserving Layer
3.1 Network Structure
3.2 Adversarial Training
3.3 Using Representative Points
4 Theoretical Analysis
5 Experiments
6 Conclusion
References
Metric Learning on the Manifold of Oriented Ellipses: Application to Facial Expression Recognition
1 Introduction
2 A Class of Metrics on the Shape Space S+c(2,n)
2.1 Gradient
3 Metric Learning
4 Experiments
4.1 Formulation of Facial Expression Recognition in S+(d,n)
4.2 Experimental Setting
4.3 Datasets
4.4 Results and Discussion
5 Conclusion
References
MANPU - The 4th International Workshop on coMics ANalysis, Processing and Understanding
The 4th International Workshop on coMics ANalysis, Processing and Understanding (MANPU2020)
Workshop Description
Organization
General Co-chairs
Program Co-chairs
Advisory Board
Program Committee
An OCR Pipeline and Semantic Text Analysis for Comics
1 Introduction: Context & Previous Work
2 Methodology and Dataset
3 OCR Pipeline: Results and Discussion
4 Some Textual Properties of 129 Graphic Novels
5 Conclusion and Future Work
References
Manga Vocabulometer, A New Support System for Extensive Reading with Japanese Manga Translated into English
1 Introduction
2 Related Work
3 Application Architecture
3.1 Each Component of Manga Vocabulometer
3.2 Learning Material
4 Experiment
4.1 Experimental Condition
4.2 Details of Experiment
5 Result and Discussion
6 Conclusion and Future Work
References
Automatic Landmark-Guided Face Image Generation for Anime Characters Using C2GAN
1 Introduction
2 Related Works
2.1 Face Image Generation for Anime Characters
2.2 Face Image Generation Based on Landmarks
3 Proposed Method
3.1 Outline of Proposed Method
3.2 Facial Landmarks of Anime Characters as Keypoints
3.3 Network Structure
3.4 Anime Character Dataset Generation
4 Experiment
4.1 Anime Character Face Image Generation Using C2GAN
4.2 Effect of Styles of Landmarks to Performance
5 Conclusion
References
Text Block Segmentation in Comic Speech Bubbles
1 Introduction
2 Related Works
3 Proposed Approach
4 Results
4.1 eBDtheque Dataset
4.2 Manga109 Dataset
4.3 Private Dataset
4.4 Synthesis
5 Conclusion
References
MMDLCA - Multi-modal Deep Learning: Challenges and Applications
Workshop on Multi-modal Deep Learning: Challenges and Applications (MMDLCA)
Workshop Description
Organization
MMDLCA Chairs
Program Committee
Additional Reviewers
Hierarchical Consistency and Refinement for Semi-supervised Medical Segmentation
1 Introduction
2 Method
2.1 Hierarchical Task Decompose
2.2 Multi-task Mean Teacher Framework
3 Experiments
3.1 Experiment Setting
3.2 Evaluation of Our Method
3.3 Comparison with the State-of-the-Art
4 Conclusion
References
BVTNet: Multi-label Multi-class Fusion of Visible and Thermal Camera for Free Space and Pedestrian Segmentation
1 Introduction
2 Literature Review
3 Algorithm
3.1 Encoder Branches
3.2 Decoder Branch
3.3 Output Branches
3.4 Training
3.5 Post-processing
4 Algorithm Variants: Ablation Study
5 Experimental Results
5.1 Comparative Analysis
5.2 Ablation Study
5.3 Architecture Variations
5.4 Semantic Segmentation Formulation
5.5 Boundary Estimation and Integration
6 Conclusion
References
Multimodal Emotion Recognition Based on Speech and Physiological Signals Using Deep Neural Networks
1 Introduction
2 Related Work
3 Proposed Method
4 Experiments and Results
4.1 Dataset
4.2 Objective Function
4.3 Experimental Setup
4.4 Experimental Results
5 Conclusion
References
Cross-modal Deep Learning Applications: Audio-Visual Retrieval
1 Introduction
2 Related Work
3 Proposed Models
3.1 Feature Extraction
3.2 Embedding Network
3.3 Training Loss Function
4 Experiment
4.1 Datasets and Evaluation Metrics
4.2 Experiment Setting
4.3 Experiment Result and Analysis
5 Conclusion
References
Exploiting Word Embeddings for Recognition of Previously Unseen Objects
1 Introduction
2 Motivation
3 Related Work
4 Description of the Proposed Approach
5 Experiments
5.1 Experiment Set 1
5.2 Experiment Set 2
6 Conclusions
References
Automated Segmentation of Lateral Ventricle in MR Images Using Multi-scale Feature Fusion Convolutional Neural Network
1 Introduction
2 Materials and Methods
2.1 Data and Preprocessing
2.2 MFF-Net
3 Experimental Results and Discussion
3.1 Parameter Setting
3.2 Evaluation Metrics
3.3 Results and Discussion
4 Conclusion
References
Visual Word Embedding for Text Classification
1 Introduction
2 Related Work
3 Proposed Approach
3.1 Encoding Scheme
3.2 Encoding Scheme with CNN
4 Dataset
5 Experiments
5.1 Parameters Setting
5.2 Data Augmentation
5.3 Comparison with Other State-of-the-art Text Classification Methods
5.4 Comparison with State-of-the-Art CNNs
6 Multimodal Application
7 Conclusion
References
CC-LSTM: Cross and Conditional Long-Short Time Memory for Video Captioning
1 Introduction
2 Related Work
3 The Proposed Approach
3.1 Problem Definition
3.2 CC-LSTM for Video Captioning
4 Experiments
4.1 Implementation Details
4.2 Experiment Results on the MSVD Dataset
4.3 Experiment Results on the MSR-VTT Dataset
5 Conclusion
References
An Overview of Image-to-Image Translation Using Generative Adversarial Networks
1 Introduction
2 I2I Translation Preliminaries
2.1 Generative Adversarial Networks
2.2 Normalization
2.3 Evaluation Metrics
3 I2I Translation Methods
3.1 Supervised Methods
3.2 Unsupervised Methods
4 Challenges and Future Directions
5 Conclusion
References
Fusion Models for Improved Image Captioning
1 Introduction
2 Related Work
3 Fusion Techniques and Variations
3.1 Auxiliary Language Model
3.2 Fusion Strategies and Architecture
4 Experiments
4.1 Baseline
4.2 Fusion Model Training
5 Results
5.1 Quantitative Analysis
5.2 Qualitative Analysis
6 Conclusion
A Appendix
References
From Bottom to Top: A Coordinated Feature Representation Method for Speech Recognition
1 Introduction
2 Sparse Representation of Speech Signals
3 The Proposed MSR for Speech Recognition
3.1 MFCC Aided Sparse Representation (MSR)
3.2 Neural Network Based on Attention Mechanism for Speech Recognition
4 Experiment Setup
5 Experiment Results
6 Conclusion
References
MMForWild2020 - MultiMedia FORensics in the WILD 2020
MultiMedia FORensics in the WILD (MMForWILD) 2020 ICPR Workshop - January 2021
Workshop Description
Organization
Workshop Chairs
Publication Chair
Program Committee
Industrial Sponsor
Increased-Confidence Adversarial Examples for Deep Learning Counter-Forensics
1 Introduction
2 Proposed Confidence-Controlled Attacks
3 Methodology
3.1 Attacks
3.2 Datasets and Networks
4 Experiments
4.1 Setup
4.2 Results
5 Discussion and Conclusions
References
Defending Neural ODE Image Classifiers from Adversarial Attacks with Tolerance Randomization
1 Introduction
2 Related Work
3 N-ODE Nets and Carlini and Wagner Attack
3.1 Neural ODE Networks
3.2 The Carlini and Wagner Attack
4 The Proposed Decision Method Based on Tolerance Randomization
4.1 On Tolerance Variation
4.2 Tolerance Randomization to Detect Adversarial Samples
5 Experimental Setup
5.1 Datasets: MNIST and CIFAR-10
5.2 Details on Training
5.3 Carlini and Wagner Attack Implementation Details
6 Experimental Results
6.1 Results Varying the Tolerance at Test-Time
6.2 Results on Detection of Adversarial Samples
7 Conclusions and Future Work
References
Analysis of the Scalability of a Deep-Learning Network for Steganography ``Into the Wild''
1 Introduction
2 Model Scaling and Data Scaling
3 A Test Bench to Assess Scalability for DL-based Steganalysis
3.1 Discussion on the Test Bench Design
3.2 Presentation of LC-Net
4 Experiments and Results
4.1 Dataset and Software Platform
4.2 Training, Validation, and Testing
4.3 Hyper-parameters
4.4 Results and Discussion
5 Conclusion
References
Forensics Through Stega Glasses: The Case of Adversarial Images
1 Introduction
2 Related Works
2.1 Steganalysis for Forensic Purposes
2.2 Adversarial Examples
2.3 Defenses
2.4 Steganographic Costs
2.5 Looking at Adversarial Examples with Stega Glasses
3 Steganographic Post-Processing
3.1 Optimal Post-processing
3.2 Our Proposal
3.3 Simplification for Quadratic Stego-Costs
4 Experimental Investigation
4.1 Experimental Setup
4.2 Robustness of Recent Classifiers: There is Free Lunch
4.3 Detection with Forensics Detectors
4.4 Post-processing with a Steganographic Embedder
4.5 Training on Adversarial Images with GINA Costs
5 Conclusions
References
LSSD: A Controlled Large JPEG Image Database for Deep-Learning-Based Steganalysis ``Into the Wild''
1 Introduction
2 A ``Controlled'' Procedure to get a JPEG image
2.1 RAW Image Sources
2.2 The ``Development'' pipeline
2.3 Development Parameters
2.4 Choice of the JPEG Quality Factor
2.5 Reflection About Quantization Matrix Diversity
3 Application to DL-Based Steganalysis
3.1 Training Database Construction
3.2 Test Database Creation
3.3 Format of Images
4 Conclusion
References
Neural Network for Denoising and Reading Degraded License Plates
1 Introduction
2 Related Works
3 Network
3.1 Denoising Network
3.2 Reading Network
4 Dataset
5 Training and Results
6 Conclusion
References
The Forchheim Image Database for Camera Identification in the Wild
1 Introduction
2 Related Work
3 The Forchheim Image Database
4 Camera Identification: Dataset Split, Methods, and Training Augmentation
4.1 Dataset Splits
4.2 Compared Methods
4.3 Matching the Network Input Resolutions
4.4 Training Augmentation
5 Results
5.1 Performance Under Ideal Conditions
5.2 Robustness Against Known Post-processing
5.3 Robustness Against Unknown Real-World Post-processing
5.4 Impact of Scene Splitting
6 Conclusion
References
Nested Attention U-Net: A Splicing Detection Method for Satellite Images
1 Introduction
2 Related Work
3 Proposed Approach
4 GAN Generated Images
5 Experiments
6 Conclusion
References
Fingerprint Adversarial Presentation Attack in the Physical Domain
1 Introduction
2 CNN-Based FPAD and Adversarial Perturbations
3 Proposed Approach
3.1 Fingerprints Liveness Dataset
3.2 Adversarial Perturbations for Fingerprints
3.3 Spoof's Creation and Acquisition
3.4 Attacking the CNN for Liveness Detection
4 Results
5 Conclusions
References
Learning to Decipher License Plates in Severely Degraded Images
1 Introduction
2 Related Work
2.1 Pipeline-Based Recognition
2.2 Learning-Based Approaches
3 Methods
3.1 Network Architecture
3.2 Synthetic Training Data
4 Experiments
4.1 Evaluation Metric
4.2 Performance on Degraded Real-World Test Data
4.3 Experiments on the Impact of Image Degradations
5 Conclusion
References
Differential Morphed Face Detection Using Deep Siamese Networks
1 Introduction
2 Related Works
3 Method
3.1 Experimental Setup
4 Results
4.1 Metrics
5 Conclusion
References
In-Depth DCT Coefficient Distribution Analysis for First Quantization Estimation
1 Introduction
2 Related Works
3 JPEG Notation
4 FQE Through Comparison
4.1 Retrieval Distributions
4.2 Quantization Factor Estimation
4.3 Regularization
5 Parameters Setting
5.1 Clustering
5.2 Regularization
6 Experimental Results
6.1 Comparison Tests
6.2 Discussion on Unpredictable Factors
6.3 Generalizing Property
7 Conclusion
References
MOI2QDN - Workshop on Metrification and Optimization of Input Image Quality in Deep Networks
Workshop on Metrification and Optimization of Input Image Quality in Deep Networks (MOI2QDN)
Workshop Description
Organization
MOI2QDN Chairs
Program Committee
Invited Speakers - Abstracts
Imaging and Metric Considerations for DNNS
Impact of Color on Deep Convolutional Neural Networks
On the Impact of Rain over Semantic Segmentation of Street Scenes
1 Introduction
2 Methodology for Rain Generation, Rain Removal, and Semantic Segmentation
2.1 Synthetic Rain Augmentation
2.2 Generative Adversarial Network for Rain Removal
2.3 Encoder-Decoder Network for Semantic Segmentation
3 Experiments
3.1 Dataset and Evaluation Metrics
3.2 Experimental Results
3.3 Visual Inspection
4 Conclusions
References
The Impact of Linear Motion Blur on the Object Recognition Efficiency of Deep Convolutional Neural Networks
1 Introduction
2 Methodology
3 Experimental Results
3.1 Qualitative Results
3.2 Quantitative Results
4 Conclusions
References
Performance of Deep Learning and Traditional Techniques in Single Image Super-Resolution of Noisy Images
1 Introduction
2 Recent Works
2.1 Deep Learning for Single Image Super-Resolution
2.2 Deep Learning for Image Denoising
2.3 FSRCNN and IRCNN
3 Experimental Design
3.1 Implemented CNN
3.2 Training, Validation and Test Datasets
3.3 Image Pre-processing
3.4 Set of Experiments
4 Results
4.1 Experiment 1.1
4.2 Experiment 1.2
4.3 Experiment 2.1
4.4 Experiment 2.2
4.5 Experiment 2.3
4.6 Experiment 3.1
4.7 Experiment 3.2
4.8 Results Summary
5 Conclusions and Further Developments
References
The Effect of Noise and Brightness on Convolutional Deep Neural Networks
1 Introduction
2 Methodology
2.1 Sensor Noise Model
2.2 Synthetic Noise Emulation
3 Experimental Results
3.1 Methods
3.2 Dataset
3.3 Parameter Selection
3.4 Qualitative Results
3.5 Quantitative Results
4 Conclusions
References
Exploring the Contributions of Low-Light Image Enhancement to Network-Based Object Detection
1 Introduction
2 Related Work
3 Method
3.1 Detection Network Preparation
3.2 Feature Maps and Activation Extraction
3.3 Feature Enhancement Statistics Generation
4 Results
4.1 Detection Performance
4.2 Statistics Generation
4.3 Qualitative Exploration
5 Conclusion
References
Multi-level Fusion Based Deep Convolutional Network for Image Quality Assessment
1 Introduction
2 Related Works
3 Proposed Multi-level Fusion Based IQA Network
3.1 Framework
3.2 Edge Feature Fusion Strategy
3.3 Conventional CNN to Depth-Wise DO-Conv
3.4 Multi-level Fusion Strategy
4 Experiments and Results
4.1 Datasets
4.2 Evaluation Metrics
4.3 In-Dataset Evaluation
4.4 Cross-dataset Evaluation
4.5 Ablation Analysis
5 Conclusion
References
CNN Based Predictor of Face Image Quality
1 Introduction
2 The Proposed Approach: CNN-FQ and Its Learning
2.1 Statistical Model of Triplet Ranking Errors
2.2 Learning Model Parameters by EM Algorithm
3 Experiments
3.1 Evaluated Face Quality Extractors
3.2 Implementation Details
3.3 Evaluation Protocol and Results
3.4 Impact of Covariates on Quality Scores
4 Conclusions
References
MPRSS - 6th IAPR Workshop on Multimodal Pattern Recognition for Social Signal Processing in Human Computer Interaction
Preface
Organization
General Chairs
Program Committee
Explainable Model Selection of a Convolutional Neural Network for Driver's Facial Emotion Identification
1 Introduction
2 Related Work
3 The Proposed Explainable Model Selection Approach
3.1 Face Extraction
3.2 CNN Model
3.3 Explainable Model
4 Experimental Results
4.1 Datasets
4.2 Discussion and Analysis
5 Conclusion
References
Artificial Kindness The Italian Case of Google Mini Recognition Patterns
1 Introduction
2 From Human To Artificial Kindness Model
3 "Sorry, I Try My Best": An Exploratory Study of a Pattern Recognition of Artificial Kindness
3.1 Hypothesis and Methodology
4 Results
5 Conclusion
References
Fingerspelling Recognition with Two-Steps Cascade Process of Spotting and Classification
1 Introduction
2 Proposed Framework
2.1 Problem Setting
2.2 Spotting of a Fingerspelling Sequence by TRCCA
2.3 Recognition of a Spotted Fingerspelling Sequence by OMSM
2.4 The Detailed Procedure of the Fingerspelling Recognition Framework
3 Experiments
3.1 Details of the Fingerspelling Dataset
3.2 Experimental Protocol
3.3 Experimental Results
4 Conclusions
References
CNN Depression Severity Level Estimation from Upper Body vs. Face-Only Images
1 Introduction
2 Background-Related Work
2.1 The Emergence of Automatic Depression Analysis
2.2 The Role of a CNN in Detecting Depression Symptoms from Videos
2.3 From a Classification to a Regression Problem
2.4 Previous Work on Multiple Depression Datasets
3 Datasets
3.1 The Black Dog Dataset
3.2 The Audio Visual Challenge 2013 (AVEC2013) Dataset
3.3 Conversion Scales
4 Method-Experimental Design
4.1 Data Preprocessing
4.2 Models
4.3 Model Platform and Training Parameters
4.4 Datasets Split
4.5 Data Post-processing
5 Results and Discussion
6 Conclusion
References
Range-Doppler Hand Gesture Recognition Using Deep Residual-3DCNN with Transformer Network
1 Introduction
1.1 Problem Statement and Contribution
2 Literature Survey
3 Network Architecture
3.1 Residual 3DCNN (Res3D)
3.2 Transformer Encoder Network (TENet)
4 Experiment and Discussion
4.1 Data Set
4.2 Training
4.3 Evaluation Using 50:50 Training and Testing Split
4.4 Evaluation Using Cross Validation
4.5 Performance Analysis
5 Conclusion
References
Introducing Bidirectional Ordinal Classifier Cascades Based on a Pain Intensity Recognition Scenario
1 Introduction
2 Formalisation and Related Work
2.1 Formalisation
2.2 Error Correcting Output Codes
2.3 Ordinal Classifier Cascade Architectures
2.4 Differences Between OCC Architectures and ECOC Models
3 Bidirectional OCC Architectures
4 The BioVid Heat Pain Database
4.1 Data Set Description
4.2 Feature Extraction
4.3 Recent Surveys on Machine Learning-Based Pain Assessment
5 Results and Discussion
5.1 Experimental Settings
5.2 Evaluation of Classifier Cascades: The Direction Matters
5.3 Evaluation of Imbalanced ECOC Models
5.4 Evaluation of the 1vs1 and bOCC Models
5.5 Comparison of All Models
5.6 Discussion
References
Personalized k-fold Cross-Validation Analysis with Transfer from Phasic to Tonic Pain Recognition on X-ITE Pain Database
1 Introduction
2 Data Description
2.1 Participants
2.2 Experiment Description
2.3 Data Features
2.4 Physiological Responses
2.5 Data Selection
3 Experiments and Results
3.1 Random Forest
3.2 Dense Neural Network
4 Transfer from Phasic to Tonic
5 Discussion
5.1 Interpreting Results
5.2 Added Value
5.3 Limitations
6 Conclusion
References
ODANet: Online Deep Appearance Network for Identity-Consistent Multi-person Tracking
1 Introduction
2 Related Work
3 Joint Tracking and Appearance Modeling
3.1 Variational Multiple Object Tracking in a Nutshell
3.2 Deep Probabilistic Appearance Model
3.3 Unsupervised Deep Metric Learning
4 Overall Tracking System
4.1 Deep Appearance Model Update
4.2 Birth and Visibility Processes
4.3 Implementation and Training Details
5 Experiments
5.1 Quantitative Evaluation
5.2 Qualitative Results
6 Conclusion
References
Author Index

Systemvoraussetzungen

Als PDF speichern Als Link merken