Python notebooks to classify land cover crops of satellite images from EuroSat dataset in Python 3.
The task has been done in two main approaches to evaluate their performance: classification using shallow learning techniques and using deep learning.
For the shallow learning approach, a local and global feature has been done and also a feature selection using a ranking technique. For the classification I´ve used Random Forest with high variance decision trees. For the deep learning I´ve used the raw RGB images and the ResNet50 architecture with transfer learning as suggested in Helber, et al. 2019.
The code is developed in python 3, you can run it in Colab or in your local Anaconda Enviroment.If you run it in your local enviroment make sure to use your GPU and verify if the following packages are installed
pip install tensorflow==1
pip install keras==2.2
conda install numpy
conda install pandas
conda install matplotlib
conda install scikit-learn
conda install scikit-image
conda install opencv-python
Run the first .ipynb to download the dataset in your Google Drive session from the Kaggle repository.
Run the notebooks one by one and modify them to get to the results you want.
Please read CONTRIBUTING.md for details on our code of conduct, and the process for submitting pull requests to us.
- Michelle Sainos Vizuett - Coder