Rapid map creation with machine learning and earth observation data.
Examples: Cropland, Buildings, Maize
Colab notebook tutorial demonstrating data exploration, model training, and inference over small region.
Prerequisites:
- Github account
- Github access token (obtained here)
- Forked OpenMapFlow repository
- Basic Python knowledge
Inside a Github repository run:
pip install openmapflow
openmapflow generate
This generates a project for: Adding data โ Training a model โ Creating a map
Move raw labels into project:
export RAW_LABEL_DIR=$(openmapflow datapath RAW_LABELS)
mkdir RAW_LABEL_DIR/<my dataset name>
cp -r <path to my raw data files> RAW_LABEL_DIR/<my dataset name>
Add reference to data using a LabeledDataset
object in datasets.py:
datasets = [
LabeledDataset(
dataset="example_dataset",
country="Togo",
raw_labels=(
RawLabels(
filename="Togo_2019.csv",
longitude_col="longitude",
latitude_col="latitude",
class_prob=lambda df: df["crop"],
start_year=2019,
x_y_from_centroid=False,
),
),
),
...
]
Run feature creation:
earthengine authenticate # For getting new earth observation data
gcloud auth login # For getting cached earth observation data
openmapflow create-features # Initiatiates or checks progress of features creation
# May take long time depending on amount of labels in dataset
# TODO make the end more obvious
openmapflow datasets # Shows the status of datasets
dvc commit && dvc push # Push new data to data version control
git add .
git commit -m'Created new features'
git push
# Pull in latest data
dvc pull
tar -xzf $(openmapflow datapath COMPRESSED_FEATURES) -C data
export MODEL_NAME=<model_name> # Set model name
python train.py --model_name $MODEL_NAME # Train a model
python evaluate.py --model_name $MODEL_NAME # Record test metrics
dvc commit && dvc push # Push new models to data version control
git checkout -b"$MODEL_NAME"
git add .
git commit -m "$MODEL_NAME"
git push --set-upstream origin "$MODEL_NAME"
Only available through Colab. Cloud Architecture must be deployed using the deploy.yaml Github Action.