Synthetic Data: Revolutionizing the Industrial Metaverse

Contents
1 Welcome to the Age of Data
1.​1 Origins and Evolution of Record Keeping
1.​1.​1 The Advent of Digital Computers
1.​1.​2 The Advent of Corporate Database Systems
1.​1.​3 From Data Warehousing to Big Data
1.​2 Handling Multimedia Industrial Data
1.​2.​1 Industrial Multimedia Data
1.​2.​2 Dynamics of Industrial Data
1.​2.​3 Processing Multimedia Data
1.​2.​4 Need for Synthetic Industrial Data
1.​3 From Digitization to Digitalization and Digital Transformation
1.​3.​1 No Turning Back on Digital Transformation
1.​3.​2 Emerging Digital Technologies
1.​4 The Trinity of Big Data, Cloud Computing, and Artificial Intelligence
1.​4.​1 Industrial Big Data
1.​4.​2 Data Processing Using Cloud Computing
1.​4.​3 Data Analytics and Prediction Using Artificial Intelligence
1.​4.​4 Challenges with Industrial Big Data
1.​5 The Contribution of BMW Group’s Open Source APIs
References
2 Industrial Evolution Toward the Age of Imagination
2.​1 From the Industrial Age to the Information and Imagination Ages
2.​1.​1 The First and Second Industrial Revolutions
2.​1.​2 The Third Industrial Revolution and the Age of Information
2.​1.​3 The Forth Industrial Revolution Unfolding
2.​1.​4 The Fifth Industrial Revolution Toward the Age of Imagination
2.​2 With Great Technologies Comes Great Responsibility
2.​2.​1 Transforming Jobs and Creating New Opportunities
2.​2.​2 Technology Can Also Reduce Wages
2.​2.​3 Need to Integrate New Technologies with Caution
References
3 Background and Technologies
3.​1 Artificial Intelligence
3.​1.​1 Search, Planning, and Motion
3.​1.​2 Knowledge Representation
3.​1.​3 Natural Language Processing
3.​1.​4 Machine Learning
3.​1.​5 Social Intelligence
3.​1.​6 Evolutionary and Generative Computing
3.​2 Machine Learning
3.​2.​1 Supervised Learning
3.​2.​2 Unsupervised Learning
3.​2.​3 Semi-supervised Learning
3.​2.​4 Reinforcement Learning
3.​2.​5 Deep Learning
3.​3 Computer Vision
3.​4 Graphics Rendering
3.​5 3D Scanning
3.​6 Immersive Technologies
3.​6.​1 Virtual Reality
3.​6.​2 Augmented Reality
3.​7 Robotics
3.​8 Cyber-Physical Systems
3.​9 Evolution of the Web Toward Collective Knowledge and Intelligence
3.​9.​1 From Static to Intelligent Web
3.​9.​2 From Big Data to Collective Knowledge
3.​9.​3 Toward Collective Intelligence
3.​10 IoT Technologies and Semantic Interoperability​
3.​10.​1 Internet of Things
3.​10.​2 Semantic Interoperability​
3.​11 Cloud and Edge Computing
3.​12 Big Data and Data Analytics
3.​13 Digital Twin and Digital Thread
3.​14 Toward the Industrial Metaverse
References
4 How Visual Data Is Revolutionizing the Industry World
4.​1 Industrial Production Process
4.​1.​1 Design Phase
4.​1.​2 Production Phase
4.​1.​3 Testing Phase
4.​1.​4 Training Phase
4.​1.​5 Service Phase
4.​2 Industrial Application Use Cases
4.​2.​1 Visibility
4.​2.​2 Ergonomics
4.​2.​3 Packaging
4.​2.​4 Realism
4.​2.​5 Storytelling
References
5 Digital Images – The Bread and Butter of Computer Vision
5.​1 Image Representation
5.​1.​1 Raster Images Versus Vector Images
5.​1.​2 Low-Level Visual Features
5.​1.​3 High-Level Text-Based Features
5.​1.​4 Joint Word-Image Modeling
5.​1.​5 Multi-dimensional Image Feature Indexing
5.​1.​6 3D CAD Processing
5.​2 Need for Synthetic Images
5.​2.​1 Shift in Computer Vision Paradigms
5.​2.​2 Need for Synthetic Images
5.​3 Computer Vision Datasets
5.​3.​1 General Purpose Image Datasets
5.​3.​2 Industrial Image Datasets
5.​3.​3 General Purpose Synthetic Image Datasets
5.​3.​4 Industrial Synthetic Image Datasets
References
6 Creating SORDI:​ The Largest Synthetic Dataset for Industries
6.​1 Universal Scene Description
6.​2 3D Mesh Modeling
6.​3 Material Design
6.​4 Scene Rendering and Data Cleaning
6.​5 Dataset Description
6.​6 Usage for Object Recognition
6.​7 BMW Group GitHub
References
7 Toward an Industrial Robot Gym
7.​1 Creating a Digital Mockup of the Factory
7.​1.​1 System Architecture
7.​1.​2 3D Model
7.​1.​3 Mathematical Model
7.​1.​4 Rule Model
7.​2 Augmented Digital Twin
7.​2.​1 Combining Physical and Virtual Parts
7.​2.​2 AR Calibration Process
7.​2.​3 AR Data Augmentation and Control
7.​3 Virtual Quality Assurance
7.​3.​1 Level of Development
7.​3.​2 Level of Accuracy
7.​3.​3 Level of Recognizability
7.​3.​4 Combining all Three Areas
7.​4 The Case of Iw.​hub – BMW Group’s AMR
7.​4.​1 Difficulties in Manufacturing Environments
7.​4.​2 Improving Situational Awareness
7.​4.​3 Knowledge Base
7.​4.​4 Assumptions and Rules
7.​4.​5 Framework Architecture
7.​4.​6 Using SORDI for Virtual Training in the Robot Gym
7.​5 Is the Digital Twin Worth It?​
References
8 What Is Next with SORDI
8.​1 Reality Gap
8.​1.​1 In the Beginning, There Was SORDI
8.​1.​2 Visual Gap
8.​1.​3 Content Gap
8.​2 Transfer Learning:​ A Promising Solution to the Reality Gap
8.​3 Domain Adaptation
8.​4 Domain Randomization
8.​4.​1 Scene Randomization
8.​4.​2 Camera Randomization
8.​5 Real Image Obfuscation
8.​5.​1 Obfuscation Techniques
8.​5.​2 Lack of Privacy Guarantees
8.​5.​3 Obfuscation Under Privacy Attacks
8.​5.​4 Ongoing Directions
8.​6 Facilitating Real Image Labelling
8.​6.​1 Bounding Box Automated Refinement
8.​6.​2 Dynamics of Bounding Box Refinement
8.​6.​3 BMW LabelTool Lite
8.​6.​4 Ongoing Directions
8.​7 Mixing Real and Synthetic Datasets
8.​7.​1 Real Data Acquisition
8.​7.​2 Model Training
8.​8 Toward Green Manufacturing
References