Uncertainty quantification and its applications for multimodal semantic segmentation / eingereicht von Pascal Colling, M. Sc. Wuppertal, 2022
Content
- Abstract
- Acknowledgments
- Foreword
- Contents
- Contents
- 1 Introduction
- 2 Fundamentals
- 2.1 Sensor and Data Types
- 2.2 Artificial Neural Networks
- 2.3 Object Recognition
- 2.4 Uncertainty Quantification and MetaSeg
- 3 Region-Based Active Learning using Priority Maps for Image Segmentation
- 3.1 Introduction
- 3.2 Related Work
- 3.3 A new Method for Region-Based Active Learning
- 3.3.1 Acquisition with Priority Maps
- 3.3.2 Joint Priority Maps Based on Prediction Quality Estimation and Click Estimation
- 3.4 Experiments and Results
- 3.5 Discussion
- 4 False Positive Detection and Prediction Quality Estimation for Point Cloud Segmentation
- 4.1 Introduction
- 4.2 Related Work
- 4.3 A new Method for False Positive Detection and Prediction Quality Estimation
- 4.4 Experiments and Results
- 4.4.1 SemanticKITTI
- 4.4.2 NuScenes
- 4.4.3 Metric Selection
- 4.4.4 Class- and Category-wise Evaluation
- 4.4.5 Confidence Calibration
- 4.5 Discussion
- 5 HD Lane Map Creation based on Trail Map Aggregation
- 5.1 Introduction
- 5.2 Related Work
- 5.3 A new Method for HD Lane Map Generation
- 5.4 Experiments and Results
- 5.5 Discussion and Future Work
- 6 Conclusion & Outlook
- List of Figures
- List of Tables
- List of Algorithms
- List of Notations
- List of Abbreviations
- Bibliography