The Earth's surface is continually changing and it is concerning how quickly such environmental issues are intensifying as a result of unsustainable patterns of consumption and production by humans. It is clear that we are in the midst of a global climate crisis, one that will only worsen if nothing is done to stop its effects. This report's goal is to offer a technology-driven recommendation that will help educate the public and the appropriate decision-makers on the effects of the climate disaster that human activity is causing.

Using Deep Convolutional Neural Network Models and Remote Sensing Change Detection techniques, we offer a method in this paper to measure changes in geographic features seen in high resolution multi-temporal satellite data. So far, we have experimented with various image segmentation models to obtain the most accurate results and predictions. Our deliverable model will take as input satellite imagery from publicly available sources and produce segmentation masks of geographical features present in these images, such as forest cover and water bodies. These segmentation masks will then be used by our Remote Sensing Change Detection algorithms to generate a percentage value showing how the particular feature has changed over time. We are optimistic that the deliverables of this project will contribute in the development of policies to tackle the problem by assisting policymakers in understanding the relationship between unsustainable human activities and environmental deterioration.

For the GPT_RSCD notebook, install the following libraries before use: openai, numpy, opencv

For the training model notebooks, install mmcv and mmsegmentation and run the programs.

Install angular
$npm install -g @angular/cli

Install Dependencies
$npm install --save ol @planet/client

Run the frontend
$npm install $ng serve --open

It runs on http://localhost:4200/

Run backend on virtual environment

Create new env (windows)
$pip install virtualenv
$virtualenv <my_env_name>
$<my_env_name>\Scripts\activate

Install dependencies (not needed if running backend env)
$pip install django
$pip install planet
$pip install geojsonio
$pip install django-ninja
$python -m pip install django-cors-headers
$pip install numpy
$pip install pillow
$pip install boto3
$pip install matplotlib
$pip install --upgrade openai
$pip install opencv-python
$pip install torch==1.12.0 torchvision --extra-index-url https://download.pytorch.org/whl/cu113
$pip install "mmsegmentation==0.30.0"
$pip install openmim
$mim install mmcv-full==1.6.0
$mim install mmengine

Run the backend server
$python manage.py runserver

Runs on http://localhost:8000/

To check out the interface downlaod and extract the file from this link -> https://drive.google.com/file/d/1O_QBxytSSyjd4Mug9NzjbmYI20ypkJkj/view?usp=share_link
Can be run using above commnads

For the data loader APIs, install the following libraries: boto3, pillow, matplotlib, numpy, BytesIO, opencv

The AWS access key and secret access key on this repo has been deactivated due to security reasons. Please reach out at [email protected] to get the keys we are using for this project.