search

jensdebruijn / Bayesian-updating-of-hurricane-vulnerability-functions

Using rapid damage observations from social media for Bayesian updating of hurricane vulnerability functions: A case study of Hurricane Dorian
MIT License
2 stars 0 forks source link

Clarification on Labelled Data #1

Closed sandiemann closed 1 year ago

sandiemann commented 1 year ago

Hi @jensdebruijn,

Thank you so much for the research on the hurricane vulnerability. I am a master student and i was very interested with the paper published on the topic. Although, I had a question concerning the labelled data. It is mentioned in the research paper that the data was collected from videos extracting building images and the damage ratio [0–1] and building class (i.e., low-, medium, and high-quality) was estimated for each building by experts. My question is, the expert only assess the damage ratio or the building class too. My assumption is the class was labelled before based on the steps describe in Exposure section. If not, how these building were labelled into these categories?

Thanks

jensdebruijn commented 1 year ago

Thanks for this question! The experts (structural engineers) assessed both the damage ratio and building class simultaneously. This is of course not ideal. In a best-case scenario this would be based on data on the ground or imagery from before the hurricane. However, this was unfortunately not available. Hope this helps.