Search

One of the key limitations of the many existing visual tracking method is that they are built upon low-level visual features and have limited predictability power of data semantics. To effectively fill the semantic gap of visual data in visual tracking with little supervision, we propose a tracking method which constructs a robust object appearance model via learning and transferring mid-level image representations using a deep network, i.e., Network in Network (NIN). First, we design a simple yet effective method to transfer the mid-level features learned from NIN on the source tasks with large scale training data to the tracking tasks with limited training data. Then, to address the drifting problem, we simultaneously utilize the samples collected in the initial and most previous frames. Finally, a heuristic schema is used to judge whether updating the object appearance model or not. Extensive experiments show the robustness of our method. [ABSTRACT FROM AUTHOR]

Video dehazing is of crucial importance to the outdoor surveillance, and its main challenges mainly come from the spatial-temporal coherence and computational efficiency. This paper presents an efficient real-time video dehazing approach via incremental transmission learning and spatial-temporally coherent regularization, while explicitly suppressing the possible visual artifacts. First, we propose to initialize the transmission map frame-by-frame by a boundary constrained open dark channel model. Then, supposing a scene point yields highly correlated transmission values between adjacent frames, we impose a temporally coherent term to maintain the temporal consistency of consecutive transmission values, and simultaneously derive an incremental transmission adjusting term to adapt the abrupt scene depth changes between the adjacent frames. Accordingly, the highly correlated and spatial-temporally coherent transmission maps can be optimized for each video frame dehazing, whereby the flickering artifacts can be well reduced. Finally, we group each haze/haze-free images in pair and further utilize the guided joint bilateral filter to suppress the visual artifacts that possibly amplified during the dehazing process. The experimental results show that the proposed video dehazing approach is able to well preserve the spatial-temporal coherence, runs sufficiently fast, and also performs favorably compared to the state-of-the-art methods. [ABSTRACT FROM AUTHOR]

To adaptively identify the transient modal parameters for linear weakly damped structures with slow time-varying characteristics under unmeasured stationary random ambient loads, this paper proposes a novel operational modal analysis (OMA) method based on the frozen-in coefficient method and limited memory recursive principal component analysis (LMRPCA). In the modal coordinate, the random vibration response signals of mechanical weakly damped structures can be decomposed into the inner product of modal shapes and modal responses, from which the natural frequencies and damping ratios can be well acquired by single-degree-of-freedom (SDOF) identification approach such as FFT. Hence, for the OMA method based on principal component analysis (PCA), it becomes very crucial to examine the relation between the transformational matrix and the modal shapes matrix, to find the association between the principal components (PCs) matrix and the modal responses matrix, and to turn the operational modal parameter identification problem into PCA of the stationary random vibration response signals of weakly damped mechanical structures. Based on the theory of “time-freezing”, the method of frozen-in coefficient, and the assumption of “short time invariant” and “quasistationary”, the non-stationary random response signals of the weakly damped and slow linear time-varying structures (LTV) can approximately be seen as the stationary random response time series of weakly damped and linear time invariant structures (LTI) in a short interval. Thus, the adaptive identification of time-varying operational modal parameters is turned into decompositing the PCs of stationary random vibration response signals subsection of weakly damped mechanical structures after choosing an appropriate limited memory window. Finally, a three-degree-of-freedom (DOF) structure with weakly damped and slow time-varying mass is presented to illustrate this method of identification. Results show that the LMRPCA algorithm, which is robustness to Gauss measurement noise disturbances, can well identify the transient modal parameters (transient natural frequencies & transient modal shapes) for weakly damped and slow LTV structures online only from non-stationary random response signals. [ABSTRACT FROM AUTHOR]

Vision transformer (ViT) has become a leading tool in various computer vision tasks, owing to its unique self-attention mechanism that learns visual representations explicitly through cross-patch information interactions. Despite having good success, the literature seldom explores the explainability of ViT, and there is no clear picture of how the attention mechanism with respect to the correlation across comprehensive patches will impact the performance and what is the further potential. In this work, we propose a novel explainable visualization approach to analyze and interpret the crucial attention interactions among patches for ViT. Specifically, we first introduce a quantification indicator to measure the impact of patch interaction and verify such quantification on attention window design and indiscriminative patches removal. Then, we exploit the effective responsive field of each patch in ViT and devise a window-free transformer (WinfT) architecture accordingly. Extensive experiments on ImageNet demonstrate that the exquisitely designed quantitative method is shown able to facilitate ViT model learning, leading the top-1 accuracy by 4.28% at most. More remarkably, the results on downstream fine-grained recognition tasks further validate the generalization of our proposal.

Recently, self-supervised video object segmentation (VOS) has attracted much interest. However, most proxy tasks are proposed to train only a single backbone, which relies on a point-to-point correspondence strategy to propagate masks through a video sequence. Due to its simple pipeline, the performance of the single backbone paradigm is still unsatisfactory. Instead of following the previous literature, we propose our self-supervised progressive network (SSPNet) which consists of a memory retrieval module (MRM) and collaborative refinement module (CRM). The MRM can perform point-to-point correspondence and produce a propagated coarse mask for a query frame through self-supervised pixel-level and frame-level similarity learning. The CRM, which is trained via cycle consistency region tracking, aggregates the reference & query information and learns the collaborative relationship among them implicitly to refine the coarse mask. Furthermore, to learn semantic knowledge from unlabeled data, we also design two novel mask-generation strategies to provide the training data with meaningful semantic information for the CRM. Extensive experiments conducted on DAVIS-17, YouTube- VOS and SegTrack v2 demonstrate that our method surpasses the state-of-the-art self-supervised methods and narrows the gap with the fully supervised methods.

A class-agnostic tracker typically consists of three key components, i.e., its motion model, its target appearance model, and its updating strategy. However, most recent topperforming trackers mainly focus on constructing complicated appearance models and updating strategies, while using comparatively simple and heuristic motion models that may result in an inefficient search and degrade the tracking performance. To address this issue, we propose a hierarchical tracker that learns to move and track based on the combination of data-driven search at the coarse level, and coarse-to-fine verification at the fine level. At the coarse level, a data-driven motion model learned from deep recurrent reinforcement learning provides our tracker with coarse localization of an object. By formulating motion search as an action-decision problem in reinforcement learning, our tracker utilizes a recurrent convolutional neural network based deep Q-network to effectively learn data-driven searching policies. The learned motion model cannot only significantly reduce the search space, but also provide more reliable interested regions for further verifying. At the fine level, a kernelized correlation filter (KCF) based appearance model is adopted to densely yet efficiently verify a local region centered on the predicted location from the motion model. Through using of circulant matrices and fast Fourier transformation, a large number of candidate samples in the local region can be efficiently and effectively evaluated by the KCF based appearance model. Finally, a simple yet robust estimator is designed to analyze possible tracking failure. The experiments on OTB50 and OTB100 illustrate that our tracker achieves better performance than the state-of-the-art trackers.