Keywords: ImageSegmentation, Images
Need: Understanding the carbon footprint of the NHS includes, amongst other things, knowing the land usage of NHS sites and any opportunity for carbon offset through increasing green space. This project seeks to demonstrate that using satellite images we can identify green space on NHS premises versus other terrain types.
Current Knowledge/Examples & Possible Techniques/Approaches: Terrain recognition from satellite images has been demonstrated widely and used crudely in the NHS in the past. However, current implementations are inefficient and haven’t been applied to a large number of images but only small test areas. Excellent article from the ONS data science campus - Green spaces in residential gardens
Related Previous Internship Projects: Exploratory work on the data, analysis and intelligence service conducted by NHSEI.
Enables Future Work: Demonstration of the technique would showcase semantic segmentation and analysis. Expansion to include areas around NHS premises and calculation of carbon offset if X terrain type turned into green space or vice versa. Possible further work around the NHS as an anchor institute by demonstrating impact of NHS premises on surrounding terrain.
Outcome/Learning Objectives: Minimum would be a demonstration of the technique on a small set of images. Desired output would be the application of the method to England wide NHS premises and calculations of proportion of green space with a proxy for carbon offset versus other terrain types.
Datasets: OS satellite images. NHS premise postcodes from NHS-D ODS.
Desired skill set: When applying please highlight any experience around work with imaging data, computer vision and semantic segmentation, training and evaluating models, understanding of bias in training data, coding experience (including any coding in the open), any other data science experience you feel relevant.
Return to list of all available projects.