Giter Club home page Giter Club logo

ipeo_project's Introduction

Segment2Reforest

Project for the class on Image Processing for Earth Observation - EPFL, 2023

In this project, we replicate Deforestation Alerts from the Global Forest Watch Integrated Alert System with various deep learning methods with data covering the Amazon Rainforest. In doing so, we explore the impact of working with an additional NIR channel besides the optical R-G-B bands. Our best final model leverages the MSNet architecture with all four input bands (R-G-B-NIR) and reaches an F1-Score of 0.9107 and an Accuracy of 0.9164.

Using this Repository

To code is structured as a package. To use this repository, you should first create a virtual enivronment to install the required dependencies as well as the repository itself. For instance with conda, run

conda create --name project_deforestation python=3.9
conda activate project_deforestation
pip install -r requirements.txt
pip install -e .

Important: Adjust the PATH_TO_DATA variable in the config/config.py file to point to the location of the data.

Recreating predictions

To run an example inference for one image, you can use the inference.ipynb notebook located in the root directory. To do so, you must download the final model weights via this Google Drive Link and place the weights into the project's root directory.

Two examples predictions and their respective ground truth labels are shown below:

Example 1 Example 2 Example 2

Dataset

The dataset is made available through the Norway's International Climate and Forests Initiative Satellite Data Program. The present work used a curated subset for training. The labels are sourced from the Integrated Alert System from the Global Forest Watch initiative (see a short description here).

Two examples of the data can be seen below:

Example 1 Example 2

File Structure

Source Files:

  • model: Package directory for storing model implementations.
  • utils: Package directory for utility functions and helper scripts.

Notebooks:

  • A. Data Exploration
    • A1_Explore_Data.ipynb: Notebook for data exploration and testing the data loaders.
  • B. Model Training & Evaluation Notebooks
    • B0_MLE_benchmark.ipynb: Notebook for creating and evaluating the benchmark model based on Multivariate Gaussian Maximum Likelhood Estimation.
    • B1_Deeplabv3plus.ipynb: Notebook for training and running experiments with the DeepLabv3+ Model.
    • B2_MSNet.ipynb: Notebook for training and running experiments with the MSNet Model.
    • B3_UNet.ipynb: Notebook for training and running experiments with the UNet Model.

Setup & Configuration:

  • config/config.py: Configuration file to change path to data.
  • requirements.txt: File specifying the Python dependencies.
  • pyproject.toml: Project build file.

ipeo_project's People

Contributors

lsbicego avatar jpradov avatar lebronmok avatar arrangabriel avatar

Stargazers

 avatar  avatar  avatar

Watchers

 avatar

Forkers

arrangabriel

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.