arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1806.11534cs.CVcs.LGcs.RO
收藏

基于检测的多传感器三维跟踪端到端学习

End-to-end Learning of Multi-sensor 3D Tracking by Detection

Uber Advanced Technologies Group(优步先进技术集团);University of Toronto(多伦多大学)

Davi Frossard, Raquel Urtasun

摘要:

在本文中,我们提出了一种新的方法来跟踪检测,可以利用这两个相机以及激光雷达数据,以产生非常准确的三维轨迹。为了实现这一目标,我们将问题表述为可以精确求解的线性规划,并学习卷积网络以进行检测以及端到端的匹配。我们在具有挑战性的KITTI数据集中评估了我们的模型,并显示出非常有竞争力的结果。

英文摘要:

In this paper we propose a novel approach to tracking by detection that can exploit both cameras as well as LIDAR data to produce very accurate 3D trajectories. Towards this goal, we formulate the problem as a linear program that can be solved exactly, and learn convolutional networks for detection as well as matching in an end-to-end manner. We evaluate our model in the challenging KITTI dataset and show very competitive results.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑