Visual sentiment analysis is attracting more and attention with the increasing tendency to express emotions through visual contents. Recent algorithms in convolutional neural networks (CNNs) considerably advance emotion classification, which aims distinguish differences among emotional categories assigns a single dominant label each image. However, task inherently ambiguous since an image usually evokes multiple its annotation varies from person person. In this work, we address problem via...

Automatic assessment of sentiment from visual content has gained considerable attention with the increasing tendency expressing opinions via images and videos online. This paper investigates problem analysis, which involves a high-level abstraction in recognition process. While most current methods focus on improving holistic representations, we aim to utilize local information, is inspired by observation that both whole image regions convey significant information. We propose framework...

Automatic assessment of sentiment from visual content has gained considerable attention with the increasing tendency expressing opinions on-line. In this paper, we solve problem analysis using high-level abstraction in recognition process. Existing methods based on convolutional neural networks learn representations holistic image appearance. However, different regions can have a influence intended expression. This paper presents weakly supervised coupled network two branches to leverage...

Automatic assessment of sentiment from visual content has gained considerable attention with the increasing tendency expressing opinions online. In this paper, we solve problem analysis, which is challenging due to high-level abstraction in recognition process. Existing methods based on convolutional neural networks learn representations holistic image, despite fact that different image regions can have influence evoked sentiment. introduce a weakly supervised coupled network (WSCNet). Our...

Learning computational models of image aesthetics can have a substantial impact on visual art and graphic design. Although automatic assessment is challenging topic by its subjective nature, psychological studies confirmed strong correlation between layouts perceived quality. While previous state-of-the-art methods attempt to learn holistic information using deep Convolutional Neural Networks (CNNs), our approach motivated the fact that Graph Network (GCN) architecture conceivably more...

Affective image understanding has been extensively studied in the last decade since more and users express emotion via visual contents. While current algorithms based on convolutional neural networks aim to distinguish emotional categories a discrete label space, task is inherently ambiguous. This mainly because labels with same polarity (i.e., positive or negative) are highly related, which different from concrete object concepts such as cat, dog bird. To best of our knowledge, few methods...

Emotion analysis of on-line user generated textual content is important for natural language processing and social media analytics tasks. Most previous emotion approaches focus on identifying users’ emotional states from text by classifying emotions into one the finite categories, e.g., joy, surprise, anger fear. However, there exists ambiguity characteristic analysis, since a single sentence can evoke multiple with different intensities. To address this problem, we introduce distribution...

Accurate grading of skin disease severity plays a crucial role in precise treatment for patients. Acne vulgaris, the most common adolescence, can be graded by evidence-based lesion counting as well experience-based global estimation medical field. However, due to appearance similarity acne with close severity, it is challenging count and grade accurately. In this paper, we address problem image analysis via Label Distribution Learning (LDL) considering ambiguous information among severity....

Image emotion recognition attracts much attention in recent years due to its wide applications. It aims classify the emotional response of humans, where candidate categories are generally defined by specific psychological theories, such as Ekman's six basic emotions. However, with development become increasingly diverse, fine-grained, and difficult collect samples. In this paper, we investigate zero-shot learning (ZSL) problem task, which tries recognize new unseen Specifically, propose a...

An image is worth a thousand words. Many researchers have conducted extensive studies to understand visual emotions since an increasing number of users express via images and videos online. However, most existing methods based on convolutional neural networks aim retrieve classify affective in discrete label space while ignoring both the hierarchical complex nature emotions. On one hand, different from concrete isolated object concepts ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML"...

Images play a crucial role for people to express their opinions online due the increasing popularity of social networks. While an affective image retrieval system is useful obtaining visual contents with desired emotions from massive repository, abstract and subjective characteristics make task challenging. To address problem, this paper introduces Attention-aware Polarity Sensitive Embedding (APSE) network learn representations in end-to-end manner. First, automatically discover model...

Analyzing and categorizing the style of visual art images, especially paintings, is gaining popularity owing to its importance in understanding appreciating art. The evolution painting both continuous, a sense that new styles may inherit, develop or even mutate from their predecessors multi-modal because various issues such as appearance, birthplace, origin time movement. Motivated by this peculiarity, we introduce novel knowledge distilling strategy assist feature learning convolutional...

With the popularity of social media, an increasing number people are accustomed to expressing their feelings and emotions online using images videos. An emotion-based image retrieval (EBIR) system is useful for obtaining visual contents with desired from a massive repository. Existing EBIR methods mainly focus on modeling global characteristics content without considering crucial role informative regions interest in conveying emotions. Further, they ignore hierarchical relationships between...

Visual sentiment analysis is attracting increasing attention with the rapidly growing amount of images uploaded to social networks. Learning rich visual representations often requires training deep convolutional neural networks (CNNs) on massive manually labeled data, which expensive or scarce especially for a subjective task like analysis. Meanwhile, large quantity quite available yet noisy by querying using categories as keywords, where various types related specific can be easily...

Abstract Learning discriminative representations with deep neural networks often relies on massive labeled data, which is expensive and difficult to obtain in many real scenarios. As an alternative, self-supervised learning that leverages input itself as supervision strongly preferred for its soaring performance visual representation learning. This paper introduces a contrastive framework generalizable the synthetic data can be obtained easily complete controllability. Specifically, we...

Recent years have witnessed the increasing popularity of learning-based methods to enhance color and tone images. Although these achieve satisfying performance on static images, it is non-trivial extend such image-to-image handle videos. A straight extension would easily lead computation inefficiency or distracting flickering effects. In this paper, we propose a novel image-to-video model enforcing temporal stability for real-time video enhancement, which trained using only Specifically,...