Low-light image enhancement is important for high-quality display and other visual applications. However, it a challenging task as the expected to improve visibility of an while keeping its naturalness. Retinex-based methods have well been recognized representative technique this task, but they still following limitations. First, due less-effective decomposition or strong imaging noise, various artifacts can be brought into enhanced results. Second, although priori information explored...

Many graph-based semi-supervised learning methods for large datasets have been proposed to cope with the rapidly increasing size of data, such as Anchor Graph Regularization (AGR). This model builds a regularization framework by exploring underlying structure whole dataset both datapoints and anchors. Nevertheless, AGR still has limitations in its two components: (1) anchor graph construction, estimation local weights between each datapoint neighboring anchors could be biased relatively...

Several models have been proposed to cope with the rapidly increasing size of data, such as Anchor Graph Regularization (AGR). The AGR approach significantly accelerates graph-based learning by exploring a set anchors. However, when dataset becomes much larger, still faces big graph which brings dramatically computational costs. To overcome this issue, we propose novel Hierarchical (HAGR) multiple-layer anchors pyramid-style structure. In HAGR, labels datapoints are inferred from coarsest...

Machine learning can provide deep insights into data, allowing machines to make high-quality predictions and having been widely used in real-world applications, such as text mining, visual classification, recommender systems. However, most sophisticated machine approaches suffer from huge time costs when operating on large-scale data. This issue calls for the need of Large-scale Learning (LML), which aims learn patterns big data with comparable performance efficiently. In this paper, we...

Visual and audio signals often coexist in natural environments, forming audio-visual events (AVEs). Given a video, we aim to localize video segments containing an AVE identify its category. In order learn discriminative features for classifier, it is pivotal the helpful (or positive) segment pairs while filtering out irrelevant ones, regardless whether they are synchronized or not. To this end, propose new positive sample propagation (PSP) module discover exploit closely related by...

Electroencephalogram (EEG) has been widely used in neurological disease detection, i.e., major depressive disorder (MDD). Recently, some deep EEG-based MDD detection attempts have proposed and achieved promising performance. These works, however, still suffer from the following limitations, such as insufficient exploration of topological structure, information loss caused by high-dimensional data compression, under-estimation intra-class difference inter-class similarity. To solve these...

Major depressive disorder (MDD) is one of the most common and severe mental illnesses, posing a huge burden on society families. Recently, some multimodal methods have been proposed to learn embedding for MDD detection achieved promising performance. However, these ignore heterogeneity/homogeneity among various modalities. Besides, earlier attempts interclass separability intraclass compactness. Inspired by above observations, we propose graph neural network (GNN)-based fusion strategy named...

Nowadays, vision-based computing tasks play an important role in various real-world applications. However, many vision tasks, e.g., semantic segmentation, are usually computationally expensive, posing a challenge to the systems that resource-constrained but require fast response speed. Therefore, it is valuable develop accurate and real-time processing models only limited computational resources. To this end, we propose spatial-detail guided context propagation network (SGCPNet) for...

The visual quality of photographs taken under imperfect lightness conditions can be degenerated by multiple factors, e.g., low lightness, imaging noise, color distortion, and so on. Current low-light image enhancement models focus on the improvement only, or simply deal with all degeneration factors as a whole, therefore leading to sub-optimal results. In this article, we propose decouple model into two sequential stages. first stage focuses improving scene visibility based pixel-wise...

We study the problem of active learning for multi-class classification on large-scale datasets. In this setting, existing approaches built upon uncertainty measures are ineffective discovering unknown regions, and those based expected error reduction inefficient owing to their huge time costs. To overcome above issues, paper proposes a novel query selection criterion called approximated (AER). AER, each candidate is estimated an impact over all datapoints ratio between its nearby datapoints....

Depression has a large impact on one's personal life, especially during the COVID-19 pandemic. People have been trying to develop reliable methods for depression detection task. Recently, based deep learning attracted much attention from research community. However, they still face challenge that data collection and annotation are difficult expensive. In many real-world applications, only small number of or even no training available. this context, we propose Prompt-based Topic-modeling...

Metric-based few-shot learning categorizes unseen query instances by measuring their distance to the categories appearing in given support set. To facilitate measurement, prototypes are used approximate representations of categories. However, we find prototypical generally not discriminative enough represent discrepancy inter-categorical distribution queries, thereby limiting classification accuracy. overcome this issue, propose a new <bold xmlns:mml="http://www.w3.org/1998/Math/MathML"...

Depression is one of the most prevalent mental disorders, which seriously affects one's life. Traditional depression diagnostics commonly depend on rating with scales, can be labor-intensive and subjective. In this context, automatic detection (ADD), aiming to assist medical experts in their diagnosis analysis, has been attracting more attention for its better objectivity fewer laborious interventions. A typical ADD model detects via automatically extracting task-specific features from...

Major Depressive Disorder (MDD) detection with cross-domain datasets is a crucial yet challenging application due to the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">data scarcity</i> and xmlns:xlink="http://www.w3.org/1999/xlink">isolated data island</i> issues in multimedia computing research. Given domain shift issue MDD continuous stream of incoming clinical settings, Semi-supervised Domain Adaptation (SDA) suitable for addressing...

Recently, significant progress has been made in pixel-level semantic segmentation using deep neural networks. However, for the current methods, it is still challenging to achieve balance between accuracy and computational cost. To address this issue, we propose Contextual Attention Refinement Network (CARNet). In method, construct Module (CARModule), which learns an attention vector guide fusion of low-level high-level features obtaining higher accuracy. The CARModule lightweight can be...