Subjective Assessments of Urban Neighbourhoods using Street Imagery: A Vision for an AI-based Approach 

Speakers:
Suvodeep Mazumdar & Sneha Roychowdhury

Agenda: 
13:00 - Refreshments & Networking 
13:20 - Talk
13:40 - Q&A's

Abstract:

The planning process in the UK is a highly complex system, developed over many decades, and is in the process of rapid transitions into digital planning. Among these transformations is a desire to move from an outputs-based assessment to an outcomes-based assessment process. An outcomes-based assessment process explores how qualitative (and often subjective) factors such as quality of life, sense of belonging, sense of security, attitude of communities, aesthetic value, comfort etc. can be assessed within the planning system. While traditionally such factors are assessed using survey and consultation, such methods are expensive, time consuming and often only involve a small number of residents, potentially biased toward specific demographics. Our research aims to explore the opportunistic use of street view imagery (such as Google Street View) to assess these factors - this might appear to be an unusual choice, however, anecdotally, we often find ourselves visiting unfamiliar neighbourhoods for better orientation of routes, surveying locations before buying or renting a property, looking at local shops and amenities or even surveying for parking spaces – to get a ‘feel’ for the space. 

With this broad vision, in this talk we will present two early activities that are early steps toward achieving this goal. We will initially present the findings from a literature review which explored existing literature and assessment frameworks to identify key factors for outcomes-based assessments. We will then present findings from a larger scale experiment that explored the correlation between residents' subjective assessments (of 10 urban neighbourhoods in Sheffield), obtained through virtual walkthroughs, and measures of deprivation. We invited residents in the UK through Prolific, a crowdsourcing platform and we will report on our analysis of 1135 responses. Our results reveal a nuanced relationship, suggesting potential subjective biases influencing residents' perceptions. Despite these complexities, the study highlights the value of virtual walkthroughs in offering a holistic overview of neighbourhoods. We will then present our plans for future research on taking these two studies forward, in developing AI approaches towards assessing subjective factors.