weecology / NeonTreeEvaluation_package

R package for evaluating individual tree crown predictions against a diverse benchmark dataset
22 stars 7 forks source link

Regarding training images resolution #28

Closed malayjoshi13 closed 3 years ago

malayjoshi13 commented 3 years ago

Hi, I wanted to re-train the existing model used in this package. For that, I understood that training images must be RGB and of size 400*400. But I didn't understand how to determine spatial resolution while capturing images from Google Earth. I read in the documentation that 0.1m spatial resolution is standard, so how will I know that image I am about to save from Google Earth is of 0.1m spatial resolution.

bw4sz commented 3 years ago

Sorry I missed this, this repo is for evaluation. If this is still a need post on the data repo https://github.com/weecology/NeonTreeEvaluation, or probably better yet deepforest.readthedocs.io