CALIOPE-student

A convolutional neural network model based solution that learns from the CALIOPE Urban data so as to extrapolate the levels of NO2 to other points in Catalonia

Comment

Inspiration

Based on the concepts learnt throughout our degree, we thought that a model that learnt the relationships among the data and the already high resolution estimations performed by the BSC in the CALIOPE Urban project could have a chance extrapolating these values. A convolutional neural network was our choice, ad by treating the data as a 2d image with different channels (land use, height, other contaminants' values...), it would be able to better catch the relationships among the variables.

However, the process was much more difficult than what we had in mind, due mainly to two reasons:

  • The data were very large, making it impossible for us to process it easily with the RAM available in our laptops.
  • Different data types were in very different formats, complicating even more the unification process.

All in all, we ended up learning map processing techniques, map data formats and we even attempted to process everything in the Marenostrum5 super computer, but everything got complicated when we attempted to build a container with the python environment needed for the task.

Built With

Share this project:

Updates