This repo is the official implementation of our paper: SLP: cropland parcel extraction based on surface, line and point. Our paper hasn't been accepted yet so we release the part of the code. The prediction process is shown in figure below.
Google cloud: https://drive.google.com/drive/folders/1f1jZwUCS4892bkne7Ob1iWmcFhj8aDIZ?usp=sharing
-
To obtain the cropland parcel(without using patch refinement)
python main.py --img ${image_path} --weights ${weights_path} --shp ${the_path_of_output_in_shapefile}
-
The next step is repair the break line
python PatchRefinement.py --lineDN ${lineDN_path} --img ${image_path} --weights ${weights_path}
-
In final, merge the results of the two steps.
python postprocess.py --img ${image_path} --shp ${the_path_of_output_in_shapefile}
PS: We will release the training code if our paper is accepted in the future.