An Integrated Approach for Traffic Scene Understanding from Monocular Cameras

Malte Oeljeklaus

doi:10.51202/9783186815125

Keywords

Search publication

Bibliographic data

Copyright year: 2021
Publication date: 28.09.2021
ISBN-Print: 978-3-18-381512-8
ISBN-Online: 978-3-18-681512-5
Publisher: VDI Verlag, Düsseldorf
Series: Verkehrstechnik/Fahrzeugtechnik
Volume: 815
Language: German
Pages: 142
Product type: Book Titles

Titelei/Inhaltsverzeichnis Partial access Pages I - XIV Download chapter (PDF)
1. Motivation No access
2. Outline and contributions No access
1. Advances in CNN architectures for image processing No access
2. Traffic scene representations from monocular cameras No access
3. Fundamental principles and general framework No access
1. Outline of the camera system and test platforms No access
2. Inferring scene points from image space measurements No access
1. General design considerations No access
2. Multi-task learning and architectural implications No access
3. Comparison and choice of the feature encoder architecture No access
1. Use and taxonomies of the traffic scene context No access
2. Recognition decoder and architecture integration No access
3. Road-topology recognition experiments No access
1. Traffic scene segmentation as dense classification No access
2. Segmentation decoder architecture and spatial priors No access
3. Experiments on drivable road area segmentation No access
1. Classification and localization of 2D bounding boxes No access
2. Auxiliary regressands and decoder architecture for spatial reconstruction No access
3. Object detection and reconstruction experiments No access
1. Multi-task decoder and architecture integration No access
2. Practical strategy for the joint training of all perceptual tasks No access
3. Experimental results and comparison No access
Summary, Conclusion, and Outlook No access Pages 97 - 99
1. Road topology dataset statistics No access
2. Technical specifications of the camera system No access
3. Single-task pre-rec curves for all road topologies No access
4. Overview of the segmentation decoder with Hadamard layer No access
5. Detailed breakdown of the single-task KITTI road segmentation results No access
6. Overview of the SSD decoder with auxiliary regressands No access
7. Dual-task Rec+Seg pre-rec curves for road topology recognition No access
8. Dual-task Rec+Det pre-rec curves for road topology recognition No access
9. Multi-task pre-rec curves for road topology recognition No access
10. Dual-task road topology confusion matrices No access
11. Detailed breakdown of the multi-task KITTI road segmentation results No access
12. Full runtime measurement data No access
Bibliography No access Pages 115 - 142