Speaker
Description
The rapid urbanisation with its plethora of complex problems ranging from population growth and climate change to social inequalities and resource consumption call for urgent need to look for solutions that can help towards efficient decision-making. Researchers are thus looking for new and dynamic sources of urban data that can help retrofit the traditional planning methodologies.
Open urban image datasets like Cityscapes Datasets (Cordts et al., 2016), The EuroCity Person Dataset (ECP) (Braun et al., 2019) and diverse datasets collected and provided by Mapillary (Neuhold et al., 2017) provide an opportunity in this regard. These datasets include data collected from cameras installed on streets, traffic lights, or buildings, which serve purposes like surveillance or supporting autonomous driving technologies. Additionally, some of this data is crowdsourced, reflecting diverse contributions from users. All those spatio-temporal datasets are annotated with metadata and partly segmented using various labels at different levels of details, enabling the classification of pedestrians, built environment features, vegetation cover, and more. While primarily used for training deep learning models, these datasets were not initially conceptualized for urban planning applications. However, they offer significant potential for assessing built environment characteristics, analyzing street-user groups, and mapping infrastructure defects.
These datasets are very different from the user-generated visual data through picture-based social media but are rather specifically gathered to understand urban scenes. In comparison to Street View imagery which may have a broader geographical coverage but lack structured annotations, these datasets provide labelling and annotations which are essential for training and evaluating machine learning models in applications like autonomous driving. Leveraging the potential of these datasets, that are being collected at an unprecedented rate and are going to only increase in the future, could be a key in understanding cities. Can the annotations help in the quantification of urban features which can be incorporated in urban analysis? What value addition can these datasets bring to urban analysis that commonly uses points of interests, population and demographic data. This contribution describes the aforementioned open datasets in detail and examines their suitability and limitations for use in planning.
While keeping in mind the ethical concerns of “anticipatory governance” (Kitchin, 2016) of using a dataset with limited geographical coverage, the research discusses the other side of the phenomenon, i.e. how can an easily accessible new form of data be repurposed to generate knowledge about the cities. To this end, their metadata is analysed for their quality, coverage and segmentation. In a further step, they are overlaid with other urban data sources for a specific urban region to assess their suitability in urban analytics.
In an era marked by the ad-hoc deployment of sensors within the urban fabric—often with or without the primary intent of supporting urban planning—it becomes essential to examine their potential for advancing urban analysis. These sensors hold the promise of generating rich, real-time data that can deepen our understanding of urban systems and expand the horizons of urban studies. Furthermore, their integration fosters interdisciplinary research, bridging fields such as urban planning, data science, environmental studies, and social sciences to develop more holistic and innovative solutions for complex urban challenges.
Funding
This research is part of the research project "Ageing Smart - Designing Spaces Intelligently", which is funded by the Carl Zeiss Foundation.
References
Braun, M., Krebs, S., Flohr, F. and Gavrila, D. M. (2019) EuroCity Persons: A Novel Benchmark for Person Detection in Traffic Scenes. IEEE Transactions on Pattern Analysis and Machine Intelligence, 41(8), pp.1844-1861.
Cordts, M., Omran, M., Ramos, S., Rehfeld, T., Enzweiler, M., Benenson, R., Franke, U., Roth, S. and Schiele, B. (2016) The Cityscapes Dataset for Semantic Urban Scene Understanding, Proc. of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
Neuhold, G., Ollmann, T., Rota Bulò, S. and Kontschieder, P. (2017) The Mapillary Vistas Dataset for Semantic Understanding of Street Scenes. International Conference on Computer Vision (ICCV).
Kitchin R (2016) The ethics of smart cities and urban science. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences
| Keywords | Repurposed Visual Data; Urban Analytics; Smart City; Open Data |
|---|---|
| Best Congress Paper Award | No |
