iNaturalist Geomodel Annotation Tool

Our iNatator tool is a web application for experts to annotate species range maps based on predictions from the Spatial Implicit Neural Representations (SINR)/iNaturalist Geomodel. This project started as part of the Data Science for the Common Good program run by the Center for Data Science at UMass Amherst. You can view a demo video of the tool on Youtube or see an overview of the approach in our poster.

Collaborators: Angela Zhu, Paula Navarrete, Sergei Pogorelov, Ozel Yilmazel

🌍 Overview

The iNaturalist team is interested in improving the predictions outputted by the Geomodel by asking experts that understand species ranges to share their knowledge and incorporate it into the Geomodel. To this end, we created a tool that allows obtaining expert feedback through annotations of species ranges in an interactive map.

You can use this code to, for a species of your choosing, show current Geomodel predictions from SINR models onto the map, annotate the species range map considering presence and absence data, save the annotation to the database, and retrieve previous annotations from the database.

Original SINR Information: Spatial Implicit Neural Representations for Global-Scale Species Mapping - ICML 2023 ICML 2023 paper Spatial Implicit Neural Representations for Global-Scale Species Mapping.

Getting Started

🐥 Installation for local development

1. Clone the repository git clone [email protected]:UMassCDS/inatator.git

2. For local development and testing, you can choose a database engine from two options, PostgreSQL or SQLite. The most important thing is to ensure the DATABASE_URL is configured appropriately for your database according to SQLAlchemy Database Engine docs. To aid in setting up the database we've provided example environment files, .env.copy and .docker.env.copy, where the environment variables are listed. You can copy them to .env or .docker.env and fill in the values.

   a. PostgreSQL: Run Postgres in a local Postgres server or Docker container (see provided docker-compose.yml). You should configure the following environment variables in your environment file (.env or .docker.env).

   • POSTGRES_DB: Name of your database
   • POSTGRES_USER: Username the server will use to connect to the database
   • POSTGRES_PASSWORD: Password the server will use to connect to the database
   • DATABASE_URL: SQLAlchemy database connection URL. This should be something like postgresql://${POSTGRES_USER}:${POSTGRES_PASSWORD}@db:PORT_NUMBER/${POSTGRES_DB}, but

   b. SQLite: Storing the databases in a simple SQLite database file is useful for development and testing, but shouldn't be used in production. You only need to configure the DATABASE_URL=sqlite://<path to desired database file>

3. Navigate to cloned project's root, run git submodule init and then git submodule update --remote --merge

4. You are all set for setting up the code.

Note: We need .env for local development, and .docker.env for docker containers. The difference between them is database url, which is caused by how docker manages networks.

Note: to update submodules with latest changes, run git submodule update --remote --merge

Note: src/backend/sinr is a submodule from UMassCDS/inatrualist-sinr, if you need to work on sinr code, follow development practices for that repository, this includes making a dedicated development branch, making PRs.

Note: If you need to switch to a particular UMassCDS/inaturalist-sinr branch and run the prototype, navigate to src/backend/sinr, use git checkout <branch-name> to switch to a branch. Now the submodule will be at a different branch.

Downloading the pretrained models

If you want to run the app locally for development purposes, download the pretrained models from here, unzip them, and place them in a folder located at
/src/backend/sinr/pretrained_models.
If you only run the app in Docker, there's no need to download the models; Docker will handle this for you inside the image.

Installing the app

1. We recommend using an isolated Python environment to avoid dependency issues. Install the Anaconda Python 3.9 distribution for your operating system from here.

2. Create a new environment and activate it:

 conda create -y --name inatator python==3.9
 conda activate inatator

3. After activating the environment, install the required packages:

pip install -r src/backend/requirements.txt && pip install -r src/backend/requirements-dev.txt

Note: If you get errors for psycopg2-binary, installing PostgreSQL can solve it. See PostgreSQL documentation for installation instructions and repeat step 3 after installing.

4. install js libraries needed for react

npm i --prefix src/frontend/

🐧 Running the iNatAtor Application

Run Application Locally

Terminal 1: Run the Database

1. Open a terminal and navigate to the main ds4cg2024-inaturalist directory
2. Launch postgres container:

docker compose up --build db

Terminal 2: Run the Backend

1. Launch another terminal window and navigate to project root.
2. Activate environment:

  conda activate inatator

3. Launch the backend:

  uvicorn src.backend.app.main:app --reload --env-file .env

Terminal 3: Run the Frontend

1. Launch another terminal window and navigate to project root.
2. Launch the frontend:

  npm run dev --prefix src/frontend/

In your web browser, open the link http://localhost:5173/

Run Application Locally with Docker Desktop

1. Install Docker if you haven't already. Open Docker Desktop, you cannot run containers or build images, if docker engine is not running
2. Now in terminal, navigate to project root. Ensure you have a local version of the .docker.env file with secrets.
3. Build and Compose Docker Images:
   Build docker images: For the first build it may take a while.

  docker compose build

To run the application:

  docker compose --env-file ./.docker.env up

4. You can access the application through the http://localhost on your browser. You can stop containers with ctrl+c or using the Docker app.

Additional Docker commands:

• Start the application again, run docker compose up
• Just build images, run docker compose build
• Build and run containers, run docker compose up --build
• Build only one service, run docker compose build <service-name>, for example docker compose build backend

Note: you don't have to initialize submodule to run docker, dockerfile will set up the submodules for you while building the image.

Working with the Database

Sometimes you want to run the application without containers, allowing you to develop things quickly. The Running the iNatAtor Application section explains how to run the application locally.

You can verify connections are working by going to localhost:8000/health.

Note: The codebase is getting bigger, therefore add database related code in src/backend/app/db, then make proper API routes in main, if it gets too big, we can resort to using API routers from fastapi.

Note: There are two environment files (.env and .docker.env) because the database url for local development and docker environments are separate.

Make sure you always update your local branch to the latest.

🙏 Acknowledgements

This project was enabled by data from the Cornell Lab of Ornithology, The International Union for the Conservation of Nature, iNaturalist, NASA, USGS, JAXA, CIESIN, and UC Merced. We are especially indebted to the iNaturalist and eBird communities for their data collection efforts. We also thank Grant Van Horn, Max Hamilton, Elijah Cole, Oisin Mac Aodha, Alex Shepard, Subhransu Maji, Sam Heinrich and Sarah Akbar for their suggestions and contributions to the design and testing process.

If you find our work useful in your research please consider citing the SINR Geomodel paper.

@inproceedings{SINR_icml23,
  title     = {{Spatial Implicit Neural Representations for Global-Scale Species Mapping}},
  author    = {Cole, Elijah and Van Horn, Grant and Lange, Christian and Shepard, Alexander and Leary, Patrick and Perona, Pietro and Loarie, Scott and Mac Aodha, Oisin},
  booktitle = {ICML},
  year = {2023}
}

📜 Disclaimer

Extreme care should be taken before making any decisions based on the outputs of models presented here. Our goal in this work is to demonstrate the promise of large-scale representation learning for species range estimation, not to provide definitive range maps. Our models are trained on biased data and have not been calibrated or validated beyond the experiments illustrated in the paper.