Satellite images classification

Classification of ships\icebergs images taken from space by Sentinel-1 satellites (Statoil/C-CORE Iceberg dataset), using C-Band radar. The problem was addressed by CNN with auxiliary inputs to the feed-forward part. Transfer knowledge techniques and different image filters for cleaning noisy data were applied as well

Used libraries

Used python vesrion: 3.5.2

Used libraries:

• matplotlib 2.0.2
• seaborn 0.8.1
• keras 2.0.7
• numpy 1.13.3
• pandas 0.20.1
• scikit-image 0.13.1
• scikit-learn 0.19.1
• scipy 0.19.0
• h5py 2.7.1

Data

Could be obtained here: https://www.kaggle.com/c/statoil-iceberg-classifier-challenge/data Both train.json.7z and test.json.7z should be unzipped to the project folder.

Code structure

• CNN models: iceberg_discrimination_cnn.ipynb
• Other supervised models: iceberg_discrimination_supervised_learning.ipynb

Models

Saved CNN models: saved_models folder