6 months ago
Note that the amount of storage and processing capability required for making neural networks for the complete set of data would be unfeasibly large. Hence, we would like to clarify a few points:
- You are not expected to process all the data! You should select some tiles from the large data set and focus on those. The number of tiles you analyze is up to you.
- The scope is to detect a minimum of one type of damage (for example, building structures that collapsed). If you can detect more than one type of damage, great. But you will not be penalized if you focus on one type only.
- We have selected a bundle of 11 tiles that we confirmed contain structure damage and have compiled them into this zip file: https://drive.google.com/drive/folders/1cO4VNNBuEXyd3yMM-qv_9P0iHd2psZC7?usp=sharing . This should help you train your neural networks!
We hope this helps!
IEEE GRSS Hackathon Organizing Team
If you have any questions about the hackathon, please post on the discussion forum or on Slack.