To download the dataset you can simply run this notebook in on your local machine or
This fellowchart shows the steps of the dataset creation process.
To add your own data to the dataset you can adjust the .xlsx files in the dataset folder, by addin a new row for each ROI. The .xlsx files are in the following format:
Each row in the .xlsx files represents an ROI. With the following columns:
-
name: The name of the ROI. This name will be used to create the folder structure.
-
year: The year of the ROI. This will be used to create the folder structure. (should be two digits e.g. 21)
-
roi: the roi in format of a list of coordinates. e.g.
[[[10,10],[10,20],[20,20],[20,10]]] -
date: start and end date of S2 search. (should be in format of
YYYY-MM-DD) -
priority_path: First check Ascending or Descending orbits.
-
check_second_priority_path: If the first priority path is not available, check the second priority path.
-
max_cloud: Maximum cloud coverage of S2 images.
-
max_snow: Maximum snow coverage of S2 images. If you are using high max cloud percentages, sentinel algorithm can falsely detect snow as cloud.
-
month_span: The inital expantion of month around the S2 mean date. For S1 search.
-
retry_days: The number of days to add to the span search if no S1 image was found.
-
train_test: Whether to use the ROI for training or testing. (should be
trainortest) -
type: Addional information about the ROI.
Make sure you have two .xlsx files. Each represneting a year.
Update: I am working on making a python module instead of the need of all above steps. But you're gonna need a google cloud account to use it. I will update this section as soon as I finish it.
