Deep Learning Theory and Applications (Communications in Computer and Information Science)

✍ Scribed by Ana Fred (editor), Carlo Sansone (editor), Kurosh Madani (editor)

• Publisher: Springer
• Year: 2023
• Tongue: English
• Leaves: 163
• Category: Library

✦ Synopsis

This book constitutes the refereed post-proceedings of the First International Conference and Second International Conference on Deep Learning Theory and Applications, DeLTA 2020 and DeLTA 2021, was held virtually due to the COVID-19 crisis on July 8-10, 2020 and July 7–9, 2021.
The 7 full papers included in this book were carefully reviewed and selected from 58 submissions. They present recent research on machine learning and artificial intelligence in real-world applications such as computer vision, information retrieval and summarization from structuredand unstructured multimodal data sources, natural language understanding andtranslation, and many other application domains.

✦ Table of Contents

Preface
Organization
Contents
Alternative Data Augmentation for Industrial Monitoring Using Adversarial Learning
1 Introduction
1.1 Motivation
1.2 Pushing the Limits of Image Segmentation
1.3 Structure
2 Related Work
3 Approach
3.1 Label-to-Image Model
3.2 Synthetic Label Generation
3.3 Finalizing the Training Data
4 Experiments and Discussion
4.1 Dataset Specifics
4.2 Experimental Setup
4.3 Pix2Pix Configuration
4.4 U-Net Configuration
4.5 WGAN Configuration
5 Evaluation
5.1 Discussion of Results
5.2 Assessment of Results and Insights
5.3 Visual Comparison and Critical Reflection
5.4 Summary
6 Conclusion and Outlook
References
Multi-stage Conditional GAN Architectures for Person-Image Generation
1 Introduction
2 Related Work
3 Classification of Pose Generation Methods
3.1 Model-Based Methods
3.2 Learning-Based Methods
4 Multi-stage Person Generation (MPG) Model
4.1 Base Model
5 Multi-stage Person Generation (MPG) Approaches
5.1 1: Three-stage Person Generation Approach
5.2 2: Two-stage Person Generation Approach
5.3 Loss Functions
6 Experimental Results
6.1 Datasets
6.2 Evaluation Metrics
6.3 Implementation Details
6.4 Quantitative Results
6.5 Qualitative Results
7 Conclusions
8 Future Scope of Research
References
Evaluating Deep Learning Models for the Automatic Inspection of Collective Protective Equipment
1 Introduction
2 Deep Learning in Industry Works
3 Methodology
3.1 The Database
3.2 Network Architectures
3.3 The Training/Test Cycle
4 Results and Discussion
4.1 Image Dataset
4.2 Trained Models Metrics
5 Conclusion and Future Works
References
Intercategorical Label Interpolation for Emotional Face Generation with Conditional Generative Adversarial Networks
1 Introduction
2 Background
3 Technical Framework for Interpolating Categorical Labels
3.1 Network Architecture
3.2 Interpolation
4 Feasibility Studies
4.1 Datasets
4.2 Methodology
4.3 Training
4.4 Computational Evaluation
5 Dimensional Face Generation
5.1 Emotion Models
5.2 Dataset
5.3 Methodology
5.4 Training
5.5 User Evaluation
5.6 Computational Evaluation
6 Discussion
7 Conclusion and Outlook
References
Forecasting the UN Sustainable Development Goals
1 Introduction
2 Literature Review
2.1 Short Time Series Forecasting
2.2 Time Series Causal Inference
2.3 Sustainable Development Goals Forecasting
3 The SDG Data Set and Associated Data Preparation
3.1 Taxonomy Generation (Pre-processing Stage 1)
3.2 Missing Value Imputation and Scaling (Pre-processing Stage 2)
4 The Extended SDG Track, Trace and Forecast (SDG-TTF) Model
5 Evaluation
6 System Operation
7 Conclusion
References
Disrupting Active Directory Attacks with Deep Learning for Organic Honeyuser Placement
1 Introduction
2 Related Work
3 Dataset
3.1 Extracting Active Directory Data
4 Graph Generation Framework
4.1 From AD to Input Matrices and Embeddings
4.2 The DAG-RNN Variational AutoEncoder
4.3 Honeyuser Attributes Generation
4.4 Implementation and Complexity
5 Experiments
5.1 Experimental Setup
5.2 Graph Reconstruction
5.3 New Nodes Generation
5.4 Evaluation of Nodes as Honeyusers
6 Results and Discussion
6.1 Results of Graph Reconstruction
6.2 Results of New Nodes Generation
6.3 Results of Evaluating Nodes as Honeyusers
6.4 Latent Space Exploration
7 Conclusion
References
Crack Detection on Brick Walls by Convolutional Neural Networks Using the Methods of Sub-dataset Generation and Matching
1 Introduction
2 Related Works
3 Proposed Method for Crack Detection
3.1 Structure of the Proposed Method
3.2 Sub-dataset Generation and Matching
4 Numerical Experiments
4.1 Dataset Preparation for CNN Learning
4.2 Experimental Design
4.3 Results
5 Discussion
6 Conclusions
References
Author Index