28 September 2018
King's PhD student analysing urban beauty reaches finals of UK-wide EPSRC competition
PhD student Sagar Joglekar, based in the Department of Informatics, has reached the finals of the EPSRC Connected Nation Pioneers competition for his work using data to understand human perceptions of urban spaces. Supervised by Dr Nishanth Sastry, Sagar is one of four finalists in the Creative Computing for the Digital Economy category.
The competition recognises exceptional research contributions towards the development of a ‘connected nation’, and was open to all second and third year UK postgraduate research students who could demonstrate the relevance of their work to one of four designated topics.
The categories are: Safe and Secure Cyber Society, Intelligent Informatics, Making Digital Technology Work for People, and Creative Computing for the Digital Economy. The final assessment takes place on Wednesday 17 October 2018 at The Defence Academy of the UK, Shrivenham, where finalists will showcase their work throughout the day.
Sagar told us:
‘I have always been interested in deriving meaning from the structure of data, driven by my background in computer vision and signals processing. Here at King’s I am working with terabytes of data to develop models along these goals, with images and networked structures. My PhD tested the hypothesis: "Can we develop machine learning systems to understand human perceptions?"’
Sagar worked with Bell Labs, based in Cambridge, with a team of mentors and collaborators, on a project to understand a) whether we can develop a deep learning model to learn human perceptions of urban spaces, and b) whether we can leverage this machine abstracted knowledge to generate actionable insights. The EPSRC competition project is the output of this project.
Daniele Quercia of Bell Labs explained the purpose of Facelift, the technology Sagar is developing:
'The aim of Facelift is to beautify the entire world, one Google Street view at a time. All of this is done by designing state-of-the-art deep-learning technologies that make it possible to beautify a street view and understand which urban elements have been added or removed during the beautification process. With further developments of these technologies, we would be more likely to systematically understand and re-create the environments people intuitively love.'
On his success, Sagar said:
‘I feel immensely honoured to be selected in the top 16 finalists. Especially after meeting the other 15 candidates, and understanding their projects, which are all high social and economic impact ideas, I am truly grateful for this opportunity. Exposure to competitions and peers like these allows me as a researcher to grow and imagine the real world potential of every research idea I pursue.’
Sagar’s supervisor Nishanth Sastry commented:
‘Sagar’s work could usher in a breakthrough in data-driven understanding of urban scenes. It allows urban planners and laypeople alike to understand concepts like urban beauty using the lens of data.’