Synthetic Data for Visual Machine Learning : : A Data-Centric Approach.

• Superior document: Linköping Studies in Science and Technology. Licentiate Thesis Series ; v.2202
• Author: Tsirikoglou, Apostolia.
• Place / Publishing House: Linköping : : Linkopings Universitet,, 2022.; {copy}2022.
• Year of Publication: 2022
• Edition: 1st ed.
• Language: English
• Series: Linköping Studies in Science and Technology. Licentiate Thesis Series
• Physical Description: 1 online resource (144 pages)

Table of Contents

Intro -- Abstract -- Populärvetenskaplig Sammanfattning -- Acknowledgments -- List of Publications -- Contributions -- Contents -- 1 Introduction -- 1.1 Visual data -- 1.2 Image synthesis -- 1.2.1 Computer graphics -- 1.2.2 Generative image modeling -- 1.3 Deep learning -- 1.3.1 Training data -- 1.4 Objectives -- 1.5 Outline -- 2 Background -- 2.1 Deep learning -- 2.1.1 Neural networks -- 2.1.2 Basic concepts -- 2.1.3 Applications -- 2.2 Computer vision -- 2.3 Digital pathology -- 3 Computer graphics -- 3.1 Modeling -- 3.1.1 Basics -- 3.1.2 Common practices -- 3.1.3 Procedural modeling -- 3.2 Rendering -- 3.2.1 Light transport theory -- 3.2.2 Light transport simulation -- 4 Generative modeling -- 4.1 Fundamentals -- 4.2 Deep generative models -- 4.3 Generative adversarial networks -- 4.3.1 Challenges -- 4.3.2 Common variants -- 5 Synthetic data for deep learning -- 5.1 Data-centric AI -- 5.1.1 Common practices -- 5.2 Data collection -- 5.2.1 Discussion -- 5.3 Data generation -- 5.3.1 Computer graphics -- 5.3.2 Generative adversarial networks -- 5.3.3 Contributions -- 5.3.4 Discussion -- 5.4 Data augmentation -- 5.4.1 Image manipulations -- 5.4.2 Deep learning approaches -- 5.4.3 Contributions -- 5.4.4 Discussion -- 6 Conclusion -- 6.1 Contributions -- 6.1.1 Data generation -- 6.1.2 Data augmentation -- 6.2 Discussion -- 7 Outlook -- Bibliography -- Papers.