In order to avoid collision with other traffic participants automated driving vehicles need to understand the scene around the ego-vehicle. Object detection as part of scene understanding remains a challenging task due to the highly variable object appearances. Object appearances can vary according to position, occlusion, illumination, etc. In this work we propose a combination of convolutional neural networks and context information to improve object detection. Context information and deep learning architectures, which are relevant for object detection, are chosen. Different approaches for integrating context information into the convolutional neural networt are discussed. The combined classifier is trained and evaluated on real scene data.
Object Detection Based on Deep Learning and Context Information
2016 ; Hannover, Deutschland
2016
Aufsatz (Konferenz)
Elektronische Ressource
Deutsch
Deep Learning Based Object Detection on Gas Cylinders
Springer Verlag | 2024
|Deep learning for crack-like object detection
TIBKAT | 2023
|