arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1908.01603cs.CV
收藏

长期跟踪中的模型衰减

Model Decay in Long-Term Tracking

University of Amsterdam(阿姆斯特丹大学);Kepler Vision Technologies(开普勒视觉技术公司)

Efstratios Gavves, Ran Tao, Deepak K. Gupta, Arnold W. M. Smeulders

摘要:

用不利的限界框预测更新跟踪器模型向学习添加了不可避免的偏差项。这个偏差项,我们称之为模型衰减,抵消了学习并导致跟踪漂移。虽然其不利影响在短期跟踪中可能看不出来,但这种偏差的长期积累最终可能导致目标的永久丢失。在本文中,我们从数学的角度来看模型偏差的问题。此外,我们简要地研究了各种来源的跟踪误差对模型衰减的影响,使用相关滤波器(ECO)和暹罗(SINT)跟踪器。基于观察和见解,我们提出了一些简单的补充,有助于减少长期跟踪中的模型衰减。该跟踪器在四个长期和一个短期跟踪基准上进行了评估,即使在30分钟长的视频中也表现出卓越的准确性和鲁棒性。

英文摘要:

Updating the tracker model with adverse bounding box predictions adds an unavoidable bias term to the learning. This bias term, which we refer to as model decay, offsets the learning and causes tracking drift. While its adverse affect might not be visible in short-term tracking, accumulation of this bias over a long-term can eventually lead to a permanent loss of the target. In this paper, we look at the problem of model bias from a mathematical perspective. Further, we briefly examine the effect of various sources of tracking error on model decay, using a correlation filter (ECO) and a Siamese (SINT) tracker. Based on observations and insights, we propose simple additions that help to reduce model decay in long-term tracking. The proposed tracker is evaluated on four long-term and one short term tracking benchmarks, demonstrating superior accuracy and robustness, even in 30 minute long videos.

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。

↑