| 초록 |
In this paper, we will be utilizing a deep learning model with camera sensors to detect objects, lanes, and road boundaries. By applying the ByteTrack technique, we ensure that there is stable tracking even when objects are occluded. Additionally, we will be designing a multi-tasking model that can perform various tasks simultaneously. This enables rapid environmental perception and efficient memory usage in autonomous driving systems. Moreover, this model allows achieving multi-task objectives with the use of a single neural network. This paper validates its viability through the Waymo Open Dataset. The results demonstrate that, in the case of image object detection, our multi-tasking model can outperform both DLT-Net and MultiNet in terms of accuracy and inference speed. Furthermore, relatively superior performance can be observed when detecting road boundaries and lanes as well. |
