asdspal
commented
3 years ago I won't be able to contribute.
Open wesleypergament opened 3 years ago
asdspal
commented
3 years ago I won't be able to contribute.
(using satellite imagery to assess damage after tornadoes)
What would a hack team for this project work towards in 2 days? Building a strong dataset of historical images from tornado disasters to train our machine learning software on in order for it to recognize the scale of damage after tornadoes. Houses can be ranked from (0-5) - 0=no damage, 5=total loss of property. There was a previous project done on assessing tornado damage expect it was specifically for tree damage or commercial building damage assessment (https://skytecllc.com/blog/industry/using-satellite-imagery-to-guide-natural-disaster-recovery-a-multi-agency-effort-to-map-tornado-damage-after-chattanooga-tornadoes). Our project would be specifically focused on homeowner damage.
Looking for people who can build image recognition/machine learning applications effectively. My knowledge is in the tornado space as I've been studying damage reports for the past year.
How would this project use Planet's data & platform? We would be utilizing historical imagery from the past few years in order to build an effective dataset to train the model on.
What obstacles or blockers do you think this project might run into? Not high-quality enough photos for us to build a strong enough training model.
And finally... pitch your idea: why should people hack on THIS project in particular? After tornado disasters, it is extremely important for first responders and insurance companies to know what areas/houses are most damaged. The more detailed our software is, the more of an impact it can have in the rebuilding efforts. I am sure these organizations would really be interested in buying our software as well if we create an extremely accurate model. I certainly think this is a really important and valuable problem to work on.
If you're interested in hacking on this project, add a 👍 reaction to this post.