Social Media Images Research Articles

INAPPROPRIATE CONTENT DETECTION IN IMAGE

The proliferation of social media platforms has led to a significant increase in the sharing of memes and other image-based content. However, some of these images may contain inappropriate or offensive text, which can be challenging to detect using traditional methods. Because the open-source existing solutions does not classify the text content in images. This project aims to develop a robust solution for detecting Not Safe for Work (NSFW) text content in images, with a particular focus on social media memes. The objectives of this project are to accurately extract text from images using optical character recognition (OCR) techniques, and to classify the extracted text as either NSFW or Safe for Work (SFW) using a fine-tuned natural language processing (NLP) model. The methodology employed in this project involves several key steps. First, the Keras OCR library is utilized to extract text from input images, as it has demonstrated superior performance compared to other OCR tools for social media images. Next, a preprocessing step is performed to align the extracted text in the correct sequence using coordinates, distance from the image origin, and the Pythagorean theorem. The pre-processed text is then passed to a fine-tuned BERT-BASE-UNCASED language model, which has been trained on a dataset of Reddit posts (NSFW and SFW) to classify the text as either NSFW or SFW. Key Words: OCR, Inappropriate, BERT, NSFW, SFW, Image, Text

A multimodal transfer learning framework for the classification of disaster-related social media images

Rapidly classifying disaster-related social media (SM) images during a catastrophe event is critical for enhancing disaster response efforts. However, the biggest challenge lies in acquiring labeled data for an ongoing (target) disaster to train supervised learning-based models, given that the labeling process is both time-consuming and costly. In this study, we address this challenge by proposing a new multimodal transfer learning framework for the real-time classification of SM images of the target disaster. The proposed framework is based on Contrastive Language-Image Pretraining (CLIP) model, jointly pretrained on a dataset of image-text pairs via contrastive learning. We propose two distinct methods to design our classification framework (1) Zero-Shot CLIP: it learns visual representations from images paired with natural language descriptions of classes. By utilizing the vision and language capabilities of CLIP, we extract meaningful features from unlabeled target disaster images and map them to semantically related textual class descriptions, enabling image classification without training on disaster-specific data. (2) Linear-Probe CLIP: it further enhances the performance and involves training a linear classifier on top of the pretrained CLIP model’s features, specifically tailored to the disaster image classification task. By optimizing the linear-probe classifier, we improve the model’s ability to discriminate between different classes and achieve higher performance without the need for labeled data of the target disaster. Both methods are evaluated on a benchmark X (formerly Twitter) dataset comprising images of seven real-world disaster events. The experimental outcomes showcase the efficacy of the proposed methods, with Linear-Probe CLIP achieving a remarkable 7% improvement in average F1-score relative to the state-of-the-art methods.

Social Media Image Construction and Communication of Sports Athletes

With the development of social media, it has become more and more a distinctive phenomenon for sports athletes to present themselves and spread sports culture through operating personal social media accounts. As the national sport of China, table tennis has long enjoyed widespread attention, and the athletes are also the focus of the public. In this regard, this paper selected eight Chinese table tennis athletes in different growth periods in a balanced manner according to the demographic approach. With the content and text analysis methods, the paper collects data and text from their social media accounts represented by Weibo, and analyzes the content and effects from aspects of sports communication, health communication, brand communication, social value communication and entertainment daily communication. After analyzing the current operation and exposed shortcomings of the personal social media of sports athletes represented by them, the paper puts forward three feasible suggestions: “build a good media image, convey positive energy actively,” “raise the awareness of media management, explore the possibilities of cross-border” and “enhance the ability of media operation, create differentiated personalities,” to contribute media power to the in-depth implementation of the "Healthy China" Initiative and the development of China's sports industry.

Credibility assessment of social media images shared during disasters

Credibility assessment of social media images shared during disasters

Images Connect Us Together: Navigating a COVID-19 Local Outbreak in China Through Social Media Images

Social media images, curated or casual, have become a crucial component of communicating situational information and emotions during health crises. Despite its prevalence and significance in informational dissemination and emotional connection, there lacks a comprehensive understanding of visual crisis communication in the aftermath of a pandemic which is characterized by uncertain local situations and emotional fatigue. To fill this gap, this work collected 345,423 crisis-related posts and 65,376 original images during the Xi'an COVID-19 local outbreak in China, and adopted a mixed-methods approach to understanding themes, goals, and strategies of crisis imagery. Image clustering captured the diversity of visual themes during the outbreak, such as text images embedding authoritative guidelines and "visual diaries" recording and sharing the quarantine life. Through text classification of the post that visuals were situated in, we found that different visual themes highly correlated with the informational and emotional goals of the post text, such as adopting text images to convey the latest policies and sharing food images to express anxiety. We further unpacked nuanced strategies of crisis image use through inductive coding, such as signifying authority and triggering empathy. We discuss the opportunities and challenges of crisis imagery and provide design implications to facilitate effective visual crisis communication.

Exif2Vec: A Framework to Ascertain Untrustworthy Crowdsourced Images Using Metadata

In the context of social media, the integrity of images is often dubious. To tackle this challenge, we introduce Exif2Vec , a novel framework specifically designed to discover modifications in social media images. The proposed framework leverages an image’s metadata to discover changes in an image. We use a service-oriented approach that considers discovery of changes in images as a service . A novel word-embedding-based approach is proposed to discover semantic inconsistencies in an image metadata that are reflective of the changes in an image. These inconsistencies are used to measure the severity of changes. The novelty of the approach resides in that it does not require the use of images to determine the underlying changes. We use a pretrained Word2Vec model to conduct experiments. The model is validated on two different fact-checked image datasets, i.e., images related to general context and a context-specific image dataset. Notably, our findings showcase the remarkable efficacy of our approach, yielding results of up to 80% accuracy. This underscores the potential of our framework.

Enhancing understanding of social media literacy to better inform prevention of body image and eating disorders

ABSTRACT Research has revealed an important role for appearance-focused, and in particular photo-based, social media in the development of body image and eating pathology. Social media literacy is a multifaceted construct involving purposeful selection and exposure to social media content and use of protective filtering (deliberately screening and interpreting information in a protective manner), as well as implementing critical skills related to understanding the unrealistic nature of social media content, limiting appearance comparisons with images, and contributing to social media (comments and images) in ways that limit the pursuit of appearance ideals. Previous work has provided partial support for these facets as related to lower levels of body image and eating concerns. However, additional conceptual and measurement work is needed to advance understanding of this protective role and how to foster social media literacy. Although data are scarce, targeting social media literacy in prevention and intervention programs may also be valuable. In this perspective piece, key features that we identify as priorities for future prevention and intervention efforts include developing interventions that target the implementation of social media skills beyond their acquisition. In addition, leveraging social media user generated content for prevention purposes would likely be useful, as would embedding micro-skill delivery within platforms.

The Link of Social Media and Brand Image On Consumer Purchase Intention the Mediating Role E-Wom

Electronic Word of Mouth (E-WOM) is a form of marketing, where people provide information and make honest recommendations to other people about brands, products and services. E-WOM is a new concept from WOM which provides transparency and easy access to information via the internet. The aims of this research was to determine the role of e-WOM in mediating the influence of social media and brand image on consumer purchase intention on micro, small and medium enterprises (MSMEs). The research respondents were customers of MSMEs culinary sector in Denpasar. The size of respondents were 100 culinary customers, chosen by accidental sampling or the respondents were chosen by chance at the area and was considered appropriate as a information source. Within the survey, there were five choices of scores for each question that can be chosen concurring to the real circumstance felt by the respondent. The data were analyzed with the SmartPLS 3.0. The research revealed social media and brand image has a positive and significant both direct effect as well mediated by E-WOM on customer purchase intention. The results of this research can be as reference for MSMEs that social media, brand image, positive e-WOM are important factors to increased purchase intention.

Rally ’Round the Mask: Congressional Social Media Images and Masking during COVID-19

Rally ’Round the Mask: Congressional Social Media Images and Masking during COVID-19

Social Media and Body Image from the Perspective of Secondary School Youth

The study aimed to map the prevalence of social media disorder in secondary school youth. We used an abbreviated version of the Social Media Disorder Scale (SMD) and the Body Shape Questionnaire (BSQ-16B). The reliability of the questionnaires used in the research, indicates that they have sufficient internal consistency. A total of 637 re-respondents participated in the study. Their ages ranged from 15 to 19 years (AM=16.63). The results of the study showed social media impairment in 11.1% of the respondents, confirming a statistically significant difference in terms of the gender of the adolescents in favour of girls, and in terms of the degree of concern about their appearance, in favour of those who are more concerned about their appearance. Furthermore, a statistically significant difference in the degree of concern about their appearance in terms of gender, was confirmed in favour of girls.

A (Meta) Picture is Worth a Thousand “Likes:” An Analysis of Engagement with Sports Network Images on Instagram

Previous studies on sports images have often focused on gender and race representation. However, with the rise of social media, sports photographers worldwide are fighting professionally produced and user-created images for audience attention. Recent scholarship has shown that stronger storytelling elements in sports photographs spur an increase in lower-level engagement metrics. The current study dives deeper into the types of images that generate key social media engagement. Researchers conducted a biometric analysis of 28 randomly sampled Instagram images from sports networks’ official feeds. Results indicate that social media images containing metacommunicative themes increase audience engagement. These findings offer insights into improving social media engagement for sports news sources.

A brief mindfulness meditation can ameliorate the effects of exposure to idealised social media images on self-esteem, mood, and body appreciation in young women: An online randomised controlled experiment

The study examined whether a 10-minute mindfulness meditation intervention could ameliorate effects of exposure to social media thinspiration and fitspiration images on women’s self-esteem, mood, and body appreciation. A total of 162 women aged 18–42 years (M = 25.94, SD = 4.38) completed an online experiment which involved viewing thinspiration and fitspiration images before random allocation to either a brief, ten-minute mindfulness meditation audio intervention or a ten-minute control audio about jujitsu. Participants completed self-reported measures of self-esteem, positive and negative mood, and body appreciation at baseline (Time 1), post-exposure to idealised social media images (Time 2), and immediately post-intervention (Time 3). Mixed, repeated-measures ANOVAs showed that scores were lower for body appreciation, self-esteem, and positive mood, and higher for negative mood, in both groups after exposure to idealised imagery. However, a brief mindfulness intervention ameliorated the negative effects of social media exposure. Specifically, self-esteem, body appreciation, and mood were higher in the mindfulness meditation group at Time 3, compared to the control group. Future interventions should explore the utility of mindfulness practices to provide long term buffering effects against such social media content, as well as targeting the idealisation of female physiques portrayed in thinspiration and fitspiration content.

How social media images of sexualized young women elicit appearance commentary from their peers and reinforce objectification

Social media platforms like Instagram enable users to share, view, and provide feedback on images, including photographs of oneself (e.g., selfies). In a 3 × 2 between-subjects online experiment, we investigated how women evaluate and react to photographs of their peers on social media and the role that feedback might play in both objectification of others and oneself. U.S. adult young women (N = 256; Mage = 20.06, SDage = 1.57) viewed social media images of sexualized peers, non-sexualized peers, or landscapes (control). Then, they provided feedback on the images via social media hashtags (#) or not (tagging vs. no tagging). Results revealed that participants who viewed sexualized peers demonstrated the highest levels of state self-objectification and were more likely to dehumanize the women in the photos. Hashtags generated by participants indicated that those who viewed sexualized peers engaged in greater appearance-related objectification, specifically related to body parts, and sexual objectification than those who viewed non-sexualized peers. In addition, generating hashtags that specifically focused on body parts heightened viewers’ state self-objectification. These findings illustrate the complexities of social media content production and consumption, particularly for young women.

Perceptions on Cultural Ecosystem Services of World Cultural Heritages in Beijing: A cross-cultural Comparison of Social Media-based Online Reviews and Images

This study introduces an innovative approach combining social media text and image comments to quantify the perceptual differences of cultural ecosystem services among various sociocultural groups. For five World Heritage Sites in Beijing, natural-language processing classifies visitor perceptions using social media comments and image data, and importance-performance analysis quantifies the differences between Chinese and international visitor perceptions of these cultural ecosystem services. For the six perceived cultural ecosystem service categories, cross-cultural and local visitors exhibit the same satisfaction trends. In particular, international tourists assign higher importance scores for a given cultural ecosystem service than Chinese tourists. The proposed methods incur a lower cost than traditional methods and acquire large volumes of broad-ranging data. Additionally, the importance of‒the performance analysis data-acquisition method is improved. Moreover, World Heritage Site sustainability can be promoted by understanding the different heritage tourism perceptions of different cultural groups. Furthermore, this study provides valuable insights for decision-makers to improve tourism management and conservation strategies.

Social Media Image and Computer Vision Method Application in Landscape Studies: A Systematic Literature Review

This study systematically reviews 55 landscape studies that use computer vision methods to interpret social media images and summarizes their spatiotemporal distribution, research themes, method trends, platform and data selection, and limitations. The results reveal that in the past six years, social media–based landscape studies, which were in an exploratory period, entered a refined and diversified phase of automatic visual analysis of images due to the rapid development of machine learning. The efficient processing of large samples of crowdsourced images while accurately interpreting image content with the help of text content and metadata will be the main topic in the next stage of research. Finally, this study proposes a development framework based on existing gaps in four aspects, namely image data, social media platforms, computer vision methods, and ethics, to provide a reference for future research.

Let’s see the whole picture: a multi-platform visual analysis of historic Mission San José on Instagram, Pinterest, Flickr, and TripAdvisor

ABSTRACT Tourists frequently research, plan, and reflect on their travel experiences online via social networking sites (SNSs). While past tourism research suggests that differences exist in the type of content shared by tourists on different SNSs, there is little research comparing the tourist-generated content (TGC) or user-generated photographs (UGPs) from multiple SNSs. To address this gap in literature, the current study utilized a visual content analysis of tourist-generated social media images from Instagram, Pinterest, Flickr, and TripAdvisor in order to compare these platforms and the messages they communicate about the heritage tourist destination of Mission San José located in San Antonio, Texas, United States. Employing a case study approach of a single tourist destination allowed for the exploration of how a destination can be visualized differently by tourists across different SNSs, how different SNSs are used in different ways by tourists, and how historic visual narratives are communicated in contemporary online spaces. Study findings establish the importance of analyzing visual content from across multiple SNSs because each site offers a unique perspective on the destination. These findings have implications for tourism marketers and researchers attempting to leverage user-generated online content to better understand perceived destination image, particularly in heritage tourism spaces.

Robust Training of Social Media Image Classification Models

Images shared on social media help crisis managers gain situational awareness and assess incurred damages, among other response tasks. As the volume and velocity of such content are typically high, real-time image classification has become an urgent need for faster disaster response. Recent advances in computer vision and deep neural networks have enabled the development of models for image classification for a number of tasks, including detecting crisis incidents, filtering irrelevant images, classifying images into specific humanitarian categories, and assessing the severity of the damage. To develop robust models, it is necessary to understand the capability of the publicly available pretrained models for these tasks, which remains to be underexplored in the crisis informatics literature. In this study, we address such limitations by investigating ten different network architectures for four different tasks using the largest publicly available datasets for these tasks. We also explore various data augmentation strategies, semisupervised techniques, and a multitask learning setup. In our extensive experiments, we achieve promising results.

BUSINESS INTELLIGENCE BASED RECURRENT NEURAL NETWORK RNN TECHNIQUES FOR SOCIAL MEDIA IMAGE CONTENT CLASSIFICATION

Social media platforms like X and Facebook generate vast amounts of image content daily, necessitating automated methods for classification and analysis. Integrating Business Intelligence (BI) with Recurrent Neural Network (RNN) techniques presents a promising approach to extract valuable insights from this data. This study proposes a methodology for social media image content classification using a hybrid architecture combining Convolutional Neural Networks (CNNs) for feature extraction and RNNs for capturing temporal dependencies. The model is trained on labeled image datasets from X and Facebook, leveraging transfer learning and data augmentation techniques. The contribution lies in the fusion of BI and deep learning techniques, offering a scalable solution for real-time image content classification on social media platforms. This approach enables businesses to streamline marketing analysis, trend detection, and content moderation tasks efficiently. Experimental results demonstrate the effectiveness of the proposed methodology, achieving high accuracy in classifying diverse image content. The model''s performance is validated through comprehensive evaluation metrics, showcasing its robustness and applicability in real-world scenarios.

Cross-modal retrieval of large-scale images in social media based on spatial distribution entropy

Cross-modal retrieval of large-scale images in social media based on spatial distribution entropy

Chaotic map-based selective image encryption via diffusion and confusion

Industrial diversification and Technology have a drastic impact on the day-to-day life and economic growth of the country. Sharing multimedia content over the network has drastically increased. The multimedia data can be just social media images or some highly confidential data. There is a high demand for secure image transmission over networks in the age of the Internet and other transmission mediums. Since images are high-volume data, care must be taken to ensure that the security does not become computationally very expensive. To overcome these problems, we propose a method to encrypt the images selectively using a chaos map. Selective encryption is the process of encrypting an image so that only a piece of it is visible while the remainder is encrypted. This decreases the computational cost of the algorithm to a great extent. Selective encryption consists of two parts, public and private. The public part of the image is visible to everyone but the private part is only visible to the authorized user. The chaotic map is used for encrypting the image and providing security to the algorithm. Confusion and diffusion are used for key generation using pixel value substitution and scrambling. We use PSNR, histogram analysis, and correlation coefficient to verify the result.