Method for Mapping Deforestation in Cerrado Based on Deep Learning Using Landsat and Sentinel Image Time Series

Jupyter Notebooks for the combination of LSTM and U-Net for the detection of deforestation in Cerrado, using Sentinel or Landsat image time series obtained from Brazil Data Cube.

Order for execution

The suggested order for the scripts execution is as follows:

• ./Scripts
  • Download
    • Download Images from BDC.ipynb
  • Create Cubes
    • Clip Files with Shapefile.ipynb
    • Create Data Cube Stacks.ipynb
    • Gap Filling.ipynb
  • LSTM
    • LSTM - Create Samples.ipynb
    • LSTM - Train.ipynb
    • LSTM - Predict.ipynb
  • UNET
    • UNET - Create Stack.ipynb
    • UNET - Create Samples.ipynb
    • UNET - Train.ipynb
    • UNET - Predict.ipynb

Contact

• Bruno M. Matosak
• bruno.matosak at inpe.br