Skip to content

Latest commit

 

History

History
60 lines (39 loc) · 2.36 KB

README.md

File metadata and controls

60 lines (39 loc) · 2.36 KB
Raw

Reproducibility

This folder contains the script, the notebooks and instructions needed to reproduce the results of Mateo-Garcia et al 2018: Multitemporal Cloud Masking in the Google Earth Engine.

Load the same patches from Biome dataset

The 2661 patches used in the study are stored in file splits_slices_biome.json the following python snippet opens this json file and loads an image patch from the Biome dataset (Foga et al 2017).

import json
from skimage.external import tifffile
import matplotlib.pyplot as plt


with open("splits_slices_biome.json","r") as f:
    slices_patches = json.load(f)

def to_slice(slice_list):
    return tuple([slice(*s) for s in slice_list])

product_name = "LC80150312014226LGN00"
patch_name = "009_006"
biome_image_path = "/path/to/biome/dataset/BC/"+product_name+"/"+product_name+"_B1.TIF"
slice_to_read = to_slice(slices_patches[product_name][patch_name])

# load from the image product_name the 500x500 slice corresponding to patch 009_006 
img = tifffile.imread(biome_image_path,memmap=True)[slice_to_read]

plt.figure(figsize=(10,10))
plt.imshow(img)
_ = plt.title(product_name+" B1 500x500 patch: "+patch_name)

Rerun the experiments

Run and download the data

The script reproducibility.py reruns the experiments. It applies the proposed methodology and download for each patch the Landsat 8 product and the ground truth from the Biome dataset (Foga et al 2017).

In case you don't need to download the data from the GEE see examples/cloudscore_different_preds.ipynb.

To run reproducibility.py you need to install the packages pydrive. The pydrive package is used to download the data from Google Drive. It expects the folder ee_ipl_uv_downloads exists in your Google Drive.

To donwnload a single patch run:

mkdir reproducibility_results
python reproducibility.py DownloadImageResults --image-index LC80290372013257LGN00 --split 013_011 --method percentile --basepath reproducibility_results

To download all the patches:

mkdir reproducibility_results
python reproducibility.py DownloadAll --method percentile --basepath reproducibility_results

Analyisis of the downloaded data

See the notebooks analyisis_reproducibility_example.ipynb and global_analyisis_reproducibility.ipynb.