BigEarthNet

This repository contains work related to the BigEarthNet dataset

Overview

The BigEarthNet dataset serves as a valuable resource for Earth Observation (EO) research, providing a large-scale Sentinel-2 satellite images dataset. The notebooks in this repository encompass various implementations, updates, and experiments performed on this dataset.

Contents

Notebooks

• Model_10K.ipynb: Notebook detailing updates and experiments on the BigEarthNet dataset with a subset of 10,000 samples.
• Test.ipynb: Notebook focusing on testing and validation procedures within the BigEarthNet project.
• ViewImages_earthpy.ipynb: Private Notebook for Viewing Sentinel 2 Images, utilizing EarthPy library for visualization.
• install_everything.ipynb: Comprehensive notebook covering the setup and installation procedures required for working with BigEarthNet data.
• model_resnet50_Adam.ipynb: Implementation of Resnet50 with Adam Optimizer for analyzing and processing BigEarthNet data.
• multi_label wiki.ipynb: Notebook demonstrating the handling of multi-label classification within the BigEarthNet dataset.
• preprocess_images_till_10K.ipynb: Notebook detailing data preprocessing steps until the first 10,000 samples.
• test_model.ipynb: Notebook dedicated to testing a specific model with the BigEarthNet dataset.

Requirements

Python Libraries and Modules:

The following Python libraries are used in this project:

• numpy
• json
• matplotlib
• PIL (Python Imaging Library)
• glob
• earthpy
• rasterio
• plotly
• torch
• torchvision
• scikit-image
• torch.nn
• torch.utils.data
• time
• os
• cv2 (OpenCV)
• copy
• sklearn.model_selection
• torchvision.utils

These libraries provide various functionalities for numerical computations, data manipulation, visualization, machine learning, image processing, and model handling.

Work Description

The notebooks within this repository encompass a wide range of tasks and experiments related to the BigEarthNet dataset:

• Data Exploration and Preprocessing: Several notebooks focus on exploring and preprocessing the dataset, ensuring it is ready for model training and analysis.
• Model Implementations: Notebooks like model_resnet50_Adam.ipynb detail specific models implemented for analyzing BigEarthNet data, such as Resnet50 using the Adam Optimizer.
• Testing and Validation: The repository includes notebooks dedicated to testing and validation procedures, ensuring the accuracy and reliability of models.

Usage

Each notebook serves a specific purpose within the BigEarthNet project. Ensure dependencies are installed as detailed in install_everything.ipynb. The notebooks can be run sequentially to understand and replicate the experiments conducted.

License

This project is licensed under the Apache License 2.0.