BorderAttention: a segmentation model for fields

In the NASA Harvest Field Boundary Detection Challenge this was the second place solution by the team HungryLearner.

ML Model Documentation

Please review the model architecture, license, applicable spatial and temporal extents and other details in the model documentation.

System Requirements

• Git client
• Docker with Compose v1.28 or newer.

Hardware Requirements

Inferencing	Training
16GB RAM	16GB RAM
	NVIDIA GPU

Get Started With Inferencing

First clone this Git repository.

Please note: this repository uses Git Large File Support (LFS) to include the model checkpoint file. Either install git lfs support for your git client, use the official Mac or Windows GitHub client to clone this repository.

⚡ Shell commands have been tested with Linux and MacOS but will differ on Windows, or depending on your environment.

git clone https://github.com/radiantearth/model_nasa_rwanda_field_boundary_competition_silver.git
cd model_nasa_rwanda_field_boundary_competition_silver/

After cloning the model repository, you can use the Docker Compose runtime files as described below.

Pull or Build the Docker Image

docker pull docker.io/radiantearth/mmodel_nasa_rwanda_field_boundary_competition_silver:1

Or build image from source:

cd docker-services/
docker build -t radiantearth/model_nasa_rwanda_field_boundary_competition_silver:1 .

Run Model to Generate New Inferences

1. Prepare your input and output data folders:

   • The data/input folder in this repository contains some placeholder files to guide you. The input data should follow the following convention. It should be placed in a directory named xxx_<tile_id>_<year>_<month>, where xxx is arbitrary and <tile_id> represents the id of the tile stored in that directory.

   Here is a sample for reference.

   data/input/nasa_rwanda_field_boundary_competition_source_test_00_2021_03
   data/input/nasa_rwanda_field_boundary_competition_source_test_00_2021_04
   data/input/nasa_rwanda_field_boundary_competition_source_test_00_2021_08
   data/input/nasa_rwanda_field_boundary_competition_source_test_00_2021_10
   data/input/nasa_rwanda_field_boundary_competition_source_test_00_2021_11
   data/input/nasa_rwanda_field_boundary_competition_source_test_00_2021_12
   data/input/nasa_rwanda_field_boundary_competition_source_test_01_2021_03
   data/input/nasa_rwanda_field_boundary_competition_source_test_01_2021_04
   data/input/nasa_rwanda_field_boundary_competition_source_test_01_2021_08
   data/input/nasa_rwanda_field_boundary_competition_source_test_01_2021_10
   data/input/nasa_rwanda_field_boundary_competition_source_test_01_2021_11
   data/input/nasa_rwanda_field_boundary_competition_source_test_01_2021_12
   

   These directories will contain tiff files for three tiles (id 00, and 01).

   • The output/ folder is where the model will write inferencing results.

2. Set INPUT_DATA and OUTPUT_DATA environment variables corresponding with your input and output folders. These commands will vary depending on operating system and command-line shell:

   # change paths to your actual input and output folders
   export INPUT_DATA="/home/my_user/model_nasa_rwanda_field_boundary_competition_silver/data/input/"
   export OUTPUT_DATA="/home/my_user/model_nasa_rwanda_field_boundary_competition_silver/data/output/"
   export MODELS_DIR="/home/my_user/model_nasa_rwanda_field_boundary_competition_silver/models"
   export WORKSPACE_DIR="/home/my_user/model_nasa_rwanda_field_boundary_competition_silver/workspace"

3. Run the appropriate Docker Compose command for your system:

   cd docker-services/
   docker compose up model_nasa_rwanda_field_boundary_competition_silver_v1

4. Wait for the docker compose to finish running, then inspect the OUTPUT_DATA folder for results.

Understanding Output Data

Please review the model output format and other technical details in the model documentation.