Abstract:While outdoor advertisements are common features within towns and cities, they may reinforce social inequalities in health. Vulnerable populations in deprived areas may have greater exposure to fast food, gambling and alcohol advertisements encouraging their consumption. Understanding who is exposed and evaluating potential policy restrictions requires a substantial manual data collection effort. To address this problem we develop a deep learning workflow to automatically extract and classify unhealthy advertisements from street-level images. We introduce the Liverpool 360 Street View (LIV360SV) dataset for evaluating our workflow. The dataset contains 25,349, 360 degree, street-level images collected via cycling with a GoPro Fusion camera, recorded Jan 14th - 18th 2020. 10,106 advertisements were identified and classified as food (1335), alcohol (217), gambling (149) and other (8405) (e.g., cars and broadband). We find evidence of social inequalities with a larger proportion of food advertisements located within deprived areas and those frequented by students. Our project presents a novel implementation for the incidental classification of street view images for identifying unhealthy advertisements, providing a means through which to identify areas that can benefit from tougher advertisement restriction policies for tackling social inequalities.

What problem does this paper attempt to address?

### Problems the Paper Attempts to Solve This paper aims to address the issue of how to automatically identify and classify unhealthy advertisements in street view images. Specifically, the researchers developed a deep learning workflow to automatically extract and classify unhealthy advertisements (such as fast food, alcohol, and gambling ads) from street view images. This method can be used to assess the exposure of different socio-economic groups to these advertisements, thereby providing data support for the formulation of stricter advertising restriction policies to reduce social inequality. ### Background and Motivation 1. **Impact of Outdoor Advertising**: - Studies have shown that certain social groups are more likely to be exposed to unhealthy products on various advertising platforms, exacerbating the problems of non-communicable diseases and social health inequalities. - Advertisements for unhealthy foods and beverages can promote poor dietary habits, with a particularly significant impact on children and adolescents. - Alcohol and gambling advertisements may also lead to adverse behaviors in adolescents, such as early drinking and increased gambling activities. 2. **Differences in Advertising Exposure**: - In impoverished areas, the quantity and type of outdoor advertisements are often greater, especially for fast food, alcohol, and gambling ads. - The presence of these advertisements may counteract public health efforts and exacerbate health inequalities. - The development of big data technologies allows advertisers to target specific populations more precisely, further intensifying this inequality. 3. **Challenges in Data Collection**: - Traditional data collection methods require a lot of manpower and time and are difficult to capture the dynamic changes in advertisements. - The lack of effective data systems to map the locations of advertisements limits the understanding of advertising exposure patterns and the evaluation of future regulatory interventions. ### Methods and Contributions 1. **Deep Learning Workflow**: - Developed a deep learning-based workflow to automatically extract and classify advertisements from street view images. - Used Convolutional Neural Networks (ConvNets) for image classification, which can efficiently handle large-scale data. 2. **Liverpool 360-Degree Street View Dataset (LIV360SV)**: - Created a dataset containing 25,349 street view images, each with GPS location information. - The dataset covers different areas of Liverpool, including the city center, north, and south regions. 3. **Socio-Economic Disparity Analysis**: - Analyzed social inequality by comparing the distribution of advertisements in areas with different socio-economic backgrounds. - Found that poorer areas and areas frequently visited by students have more unhealthy advertisements. ### Results 1. **Inequality in Advertisement Distribution**: - In impoverished areas of Liverpool, the proportion of unhealthy advertisements is significantly higher. - Fast food advertisements are particularly concentrated in poorer areas. - Gambling advertisements also have a higher proportion in certain socio-economic groups. 2. **Feasibility of Automatic Classification**: - Used the InceptionV3 network to classify the extracted advertisements, achieving good results. - Through training and testing, demonstrated the effectiveness of deep learning methods in automatically classifying advertisements. ### Conclusion By developing a deep learning workflow, this paper successfully achieved the goal of automatically extracting and classifying unhealthy advertisements from street view images. This method not only improves the efficiency of data collection but also provides strong support for the evaluation and formulation of advertising restriction policies, helping to reduce social inequality.

A deep learning approach to identify unhealthy advertisements in street view images

THE ALGORITHMIC FLOW OF HARMFUL INDUSTRIES ADVERTISING ON SOCIAL MEDIA PLATFORMS

Utilizing big data without domain knowledge impacts public health decision-making

The ALOS Dataset for Advert Localization in Outdoor Scenes

Deep-Learning-Based Real-Time Visual Pollution Detection in Urban and Textile Environments

Clustering of unhealthy outdoor advertisements around child-serving institutions: a comparison of three cities

Impact on Public Health Decision Making by Utilizing Big Data Without Domain Knowledge

Streetscape augmentation using generative adversarial networks: insights related to health and wellbeing

The linkage between the perception of neighbourhood and physical activity in Guangzhou, China: using street view imagery with deep learning techniques

Using Deep Learning to Examine the Association between the Built Environment and Neighborhood Adult Obesity Prevalence

Localizing Adverts in Outdoor Scenes

An analysis of food and beverage advertising on bus shelters in a deprived area of Northern England

Artificial intelligence–based assessment of built environment from Google Street View and coronary artery disease prevalence

StreetSurfaceVis: a dataset of crowdsourced street-level imagery annotated by road surface type and quality

Investigating foods and beverages sold and advertised in deprived urban neighbourhoods in Ghana and Kenya: a cross-sectional study

Evaluating the Significance of Outdoor Advertising from Driver's Perspective Using Computer Vision

Using machine learning to examine street green space types at a high spatial resolution: Application in Los Angeles County on socioeconomic disparities in exposure

Healthy vs. Unhealthy Food Images: Image Classification of Twitter Images

Revolutionising food advertising monitoring: a machine learning-based method for automated classification of food videos

Assessing the equity and evolution of urban visual perceptual quality with time series street view imagery

Automatic Understanding of Image and Video Advertisements