Multi-level Cues Research Articles

Concept Parser With Multimodal Graph Learning for Video Captioning

Conventional video captioning methods are either stage-wise or simple end-to-end. While the former might introduce additional noise when exploiting off-the-shelf models to provide extra information, the latter suffers from lacking high-level cues. Therefore, a more desired framework should be able to capture multi-aspects of videos consistently. To this end, we present a concept-aware and task-specific model named CAT that accounts for both low-level visual and high-level concept cues, and incorporates them effectively in an end-to-end manner. Specifically, low-level visual and high-level concept features are obtained from the video transformer and concept parser of CAT. And a concept loss is further introduced to regularize the learning process of concept parser w.r.t. generated pseudo ground truth. To combine multi-level features, a caption transformer is later introduced in CAT, where visual and concept features are the inputs and caption is its output. In particular, we make critical design choices in the caption transformer to learn to exploit these cues with a multi-modal graph. This is achieved by a graph loss that enforces effective learning of intra and inter correlations between multi-level cues. Extensive experiments on three benchmark datasets demonstrate that CAT achieves 2.3 and 0.7 improvements in the CIDEr metric on MSVD and MSR-VTT compared to the state-of-the-art method SwinBERT [1], and also achieves a competitive result on VATEX.

Saliency Detection via Absorbing Markov Chain with Multi-Level Cues

Saliency Detection via Absorbing Markov Chain with Multi-Level Cues

Dual adversarial convolutional networks with multilevel cues for pancreatic segmentation

Accurate organ segmentation is a relatively challenging subject in medical imaging, especially for the pancreas, whose morphological characteristics are subtle but variable. In this paper, a novel dual adversarial convolutional network with multilevel cues (DACN-MC) is proposed to segment the pancreas in computerized tomography (CT). DACN-MC first involves a duplex adversarial network using a conventional model for biomedical image segmentation, which ensures the veracity of the predicted probability volumes and ultimately enhances the quality of the obtained maps. Specifically, one of the adversarial networks helps the predicted maps to resemble the ground truths by importing extra guidance into the original loss functions. The other adversarial network further judges whether the obtained maps are well segmented and improves the image quality once again. Then, a multilevel cue collection module (MCCM) is introduced to gather many useful details for pancreas segmentation. In other words, we collect several sets of material formed by features from different layers and pick out a group with optimal performance for use in the ultimate algorithm. The experimental results show that dual adversarial convolutional networks together with multilevel cue collection help our proposed algorithm to achieve competitive segmentation performance, based on the results of several evaluation indexes.

A Pilot Study and Ecological Model of Smoking Cues to Inform Mobile Health Strategies for Quitting Among Low-Income Smokers.

One crucial factor that leads to disparities in smoking cessation between groups with higher and lower socioeconomic status is more prevalent socioenvironmental smoking cues in low-income communities. Little is known about how these cues influence socioeconomically disadvantaged smokers in real-world scenarios and how to design interventions, especially mobile phone-based interventions, to counteract the impacts of various types of smoking cues. We interviewed 15 current smokers living in low-income communities and scanned their neighborhoods to explore smoking-related experiences and identify multilevel cues that may trigger them to smoke. Findings suggest four major types of smoking cues influence low-income smokers-internal, habitual, social, and environmental. We propose an ecological model of smoking cues to inform the design of mobile health (mHealth) interventions for smoking cessation. We suggest that user-triggered strategies will be most useful to address internal cues; server-triggered strategies will be most suitable in changing perceived social norms of smoking and routine smoking activities to address social and habitual cues; and context-triggered strategies will be most effective for counteracting environmental cues. The pros and cons of each approach are discussed regarding their cost-effectiveness, the potential to provide personalized assistance, and scale.

Unsupervised Salient Object Detection by Aggregating Multi-Level Cues

In this paper, we present a novel method to detect salient object based on multi-level cues. First, a proposal processing scheme is developed by various object-level saliency cues to generate an initial saliency map. For the sake of more accurate object boundaries, a two-stage optimization mechanism is then proposed upon superpixel-level. Finally, the superpixel-level saliency map is further improved to construct the final saliency map by applying superpixel-to-pixel mapping. Extensive experimental results demonstrate that the proposed algorithm performs favorably against the state-of-art saliency detection methods in terms of different evaluation metrics on several benchmark datasets.

Biofunctionalized 3D Nanopillar Arrays Fostering Cell Guidance and Promoting Synapse Stability and Neuronal Activity in Networks.

A controlled geometry of in vitro neuronal networks allows investigation of the cellular mechanisms that underlie neuron-to-neuron and neuron–extracellular matrix interactions, which are essential to biomedical research. Herein, we report a selective guidance of primary hippocampal neurons by using arrays of three-dimensional vertical nanopillars (NPs) functionalized with a specific adhesion-promoting molecule—poly-dl-ornithine (PDLO). We show that 90% of neuronal cells are guided exclusively on the combinatorial PDLO/NP substrate. Moreover, we demonstrate the influence of the interplay between nanostructures and neurons on synapse formation and maturation, resulting in increased expression of postsynaptic density-95 protein and enhanced network cellular activity conferred by the endogenous c-fos expression. Successful guidance to foster synapse stability and cellular activity on multilevel cues of surface topography and chemical functionalization suggests the potential to devise technologies to control neuronal growth on nanostructures for tissue engineering, neuroprostheses, and drug development.

Women and the legitimization of (not) engaging in paid work: logics from Lebanon

PurposeThe purpose of this paper is to explore how public (i.e. culture, state, paid work) and private (i.e. household) patriarchal structures work to shape a woman’s own legitimacy judgments concerning not engaging in paid work. The authors trace the intersection and interaction of legitimacy logics at both the collective (i.e. validity) and individual (i.e. propriety) levels, thereby gaining a better contextual understanding of each woman’s perception of career opportunities and limitations.Design/methodology/approachQualitative methodology drawing from 35 semi-structured interviews with Lebanese women. A multilevel analytic framework combining the institutional structures of private and public patriarchy with the micro-processes of institutional logics is used.FindingsLegitimization of (not) engaging in paid work is often tied to patriarchal logics that favor private sphere responsibilities for women, particularly related to the relational and instrumental logics of childrearing and husband-oriented responsibilities. Women’s legitimacy judgment formation seems to be based on multilevel cues and on differential instances of evaluative vs passive judgment formation. Some appear to passively assume the legitimacy of the logics; while others more actively question these logics. The findings suggest that active questioning is often overwhelmed by the negative and harsh realities making the woman succumb to passivity and choosing not to engage in paid work.Originality/valueThis study provides: a better mapping of the individual woman’s daily cognitions concerning the legitimacy of (not) engaging in paid work; and a unique multilevel analytic framework that can serve as a useful example of contextualizing career research.

다중 단계 신호의 적응적 전파를 통한 동일 장면 영상의 이원 영역화

최근까지 단일 영상이나 동영상을 영역화하는 기법들은 다양하게 제시되어 왔으나, 유사한 장면에 대한 여러 장의 영상을 동시에 영역화하는 기법은 많지 않았다. 본 논문에서는 한 장소에서 연속적으로 촬영하였거나 전경 물체가 유사한 여러 영상들을 동일 장면 영상으로 정의하고, 이런 동일 장면 영상들을 적은 양의 사용자 입력을 통해 효과적으로 영역화하는 기법을 제안한다. 구체적으로, 사용자가 최초의 영상 한 장을 직접 영역화한 후, 그 영상의 영역화 결과와 영상의 특성을 토대로 다중 단계 신호를 적응적 가중치를 주어서 인접 영상으로 전파하고, 이를 통해 제안하는 기법은 인접 영상을 반복적으로 영역화한다. 영역화는 마르코프 랜덤 장에서의 에너지 최소화를 통해 이루어지는데, 전파되는 신호는 각 픽셀에 대한 에너지를 정의하는 바탕이 되며, 픽셀, 픽셀 패치, 그리고 영상 전체로부터 비롯되었는가에 따라 낮은 단계, 중간 단계, 그리고 높은 단계의 신호로 지칭된다. 또한 에너지 최소화 틀 안에서 전파된 신호를 통해 정의되는 에너지 역시 낮은 단계, 중간 단계, 그리고 높은 단계의 세 단계로 정의한다. 이런 과정을 통해 전파된 신호를 최대한 다양하게 활용하고, 이를 통해 다양한 영상에 영역화 결과가 일관되게 유지된다. 다양한 동일 장면 영상들에 제안하는 기법을 적용하여 성능을 평가하고, 픽셀 패치를 바탕으로 하는 중간 단계 신호만을 이용한 결과와 제안하는 다중 신호를 적용하는 기법의 결과를 비교한다. So far, many methods for segmenting single images or video have been proposed, but few methods have dealt with multiple images with analogous content. These images, which we term consistent scene images, include concurrent images of a scene and gathered images of a similar foreground, and may be collectively utilized to describe a scene or as input images for multi-view stereo. In this paper, we present a method to segment these images with minimum user input, specifically, manual segmentation of one image, by iteratively propagating information via multi-level cues with adaptive confidence depending on the nature of the images. Propagated cues are used as the bases to compute multi-level potentials in an MRF framework, and segmentation is done by energy minimization. Both cues and potentials are classified as low-, mid-, and high- levels based on whether they pertain to pixels, patches, and shapes. A major aspect of our approach is utilizing mid-level cues to compute low- and mid- level potentials, and high-level cues to compute low-, mid-, and high- level potentials, thereby making use of inherent information. Through this process, the proposed method attempts to maximize the amount of both extracted and utilized information in order to maximize the consistency of the segmentation. We demonstrate the effectiveness of the proposed method on several sets of consistent scene images and provide a comparison with results based only on mid-level cues [1].