Soil erosion detection - Quantum Task

Here is one of the open problems in Quantum. Soil erosion is a really unwanted process that spoils huge areas of fertile land. Your task is to train model for erosion detection.

What do you have

• Sentinel2 tile (T36UXV_20200406T083559_TCI_10m.jp2);
• Masks with soil erosion for this tile (masks directory);

What should we get

As usual it should be github repository that contains:

• jupyter notebook with data analysis;
• script for model training (preferably tf.keras);
• readme file;
• requirements.txt file;
• solution report. This is the most important part. We expect here not only solution description but also your proposals and result of research in different papers about soil erosion detection problem. In other words - what can help us to solve problem in the most effective way.

In this section we, by using file Preprocessing.ipynb, make analysis and preproc data

In this section we, by using file Modeling.ipynb make a U-Net model for prediction soil erosion with given satellite image.

1. Original Image: not saved here because of large size, but should be added to folder original
2. Preprocessing.ipynb: preproc data and created mask from shape file in folder mask and saved in original named label.jp2)
3. Modeling: here use all files(image mask implementation and Original Image) for modeling. As result: model.

Report created in pdf file named Improving_for_Model.pdf