摘要
由一名新闻记者-机器人与机器学习的工作人员新闻编辑每日新闻-机器学习-计算智能的最新研究结果已经发表。根据NewsRx记者在哈尔滨的新闻报道，“暹罗追踪器虽然在视觉领域越来越普遍，但在复杂的环境中，它们很容易受到语义干扰，导致特征信息的利用不足。特别是当多种干扰共同作用时，许多追踪器的性能往往会严重下降。”本研究经费来源于黑龙江省自然科学基金。为了解决上述问题，本文提出了一种鲁棒立体互感器网络，该网络采用混合连接机制，由信道特征感知网络(CFAN)、全局信道注意网络(GCAN)和多级特征增强单元(MFEU)组成。CFAN关注特定的频道信息，突出包含的目标特征，弱化语义分布特征，GCAN作为中间枢纽，主要负责建立搜索区域和模板之间的全局特征依赖关系，同时选择相关频道特征以提高模型的识别能力。最后,本文提出了一种基于变换的暹罗跟踪器(VTST),该跟踪器能较好地克服多种复杂属性的缺点,提高了多层次特征信息的提取能力.
Abstract
By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News-Current study results on Machine Learn ing - Computational Intelligence have been published. According to news reportin g from Harbin, People's Republic of China, by NewsRx journalists, research state d, "Although Siamese trackers have become increasingly prevalent in the visual t racking domain, they are easily interfered by semantic distractors in complex en vironments, which results in the underutilization of feature information. Especi ally when multiple disturbances work together, the performance of many trackers often suffers severe degradation." Financial support for this research came from Natural Science Foundation of Heil ongjiang Province. The news correspondents obtained a quote from the research from Harbin Engineeri ng University, "To solve the above problem, this paper presents a robust Stereos copic Transformer network for improving tracking performance. Using a hybrid att ention mechanism, our method is composed of a channel feature awareness network (CFAN), a global channel attention network (GCAN), and a multi-level feature enh ancement unit (MFEU). Concretely, CFAN focuses on specific channel information, while highlighting the contained target features and weakening the semantic dist ractor features. As an intermediate hub, GCAN is mainly responsible for establis hing the global feature dependencies between the search region and the template, while selecting the concerned channel features to improve the distinguishing ab ility of the model. In particular, MFEU is used to enhance multi-level feature i nformation to facilitate feature representation learning for our method. Finally , a Transformer-based Siamese tracker (named VTST) is proposed to present an eff icient tracking representation, which can gain advantages over a variety of chal lenging attributes."
NETL
NSTL