geotiff-tiler 0.1.4

Tool for creating patches from geo-referenced and non geo-referenced image and label pairs

GeoTIFF Tiler

A Python package for creating training patches from geospatial imagery and label pairs for machine learning applications.

Overview

GeoTIFF Tiler is designed to streamline the creation of training data patches from geo-referenced and non-geo-referenced image and label pairs. It helps prepare data for machine learning models requiring consistent input dimensions, particularly for geospatial applications.

Features

• Create patches of specified size from image-label pairs
• Support for various input formats:
  • Images: GeoTIFFs (geo-referenced and non-geo-referenced), STAC imagery
  • Labels: GeoTIFFs (geo-referenced and non-geo-referenced), vector data (.geojson, .gpkg, .shp)
• Intelligent patch filtering based on label content
• Padding for edge patches to maintain consistent dimensions
• Automatic handling of CRS and alignment issues
• Output in Zarr format for efficient storage and access
• Visualization tools for quality assessment

Installation

pip install geotiff-tiler

Quick Start

from geotiff_tiler.tiler import Tiler

# Define your image-label pairs with metadata
data = [{
    "image": "./path/to/image.tif",
    "label": "./path/to/label.tif",
    "metadata": {"collection": "satellite-name", "gsd": 0.5}
}]

# Initialize the tiler with your configuration
tiler = Tiler(
    input_dict=data,
    patch_size=(256, 256),  # Height, Width
    attr_field=["class", "category"],  # Field(s) in vector data to use for labels
    attr_values=[1, 2, 3],  # Values to extract from the fields
    stride=128,             # Overlap between patches
    discard_empty=True,     # Skip patches with no labels
    label_threshold=0.05,   # Minimum non-zero label coverage
    output_dir='./output/patches'
)

# Create the patches
tiler.create_tiles()

Using STAC Items

The library supports STAC (SpatioTemporal Asset Catalog) items, making it compatible with cloud-native geospatial workflows.

Parameters

• input_dict: List of dictionaries with "image", "label", and "metadata" keys
• patch_size: Tuple of (height, width) for the output patches
• attr_field: Field name(s) in vector data to use for labeling (list of strings)
• attr_values: Values to extract from the attribute field (list of strings or numbers)
• stride: Spacing between patches (determines overlap); if None, uses max(patch_size)
• discard_empty: Whether to skip patches with no labels
• label_threshold: Minimum fraction of non-zero pixels required in a label patch
• output_dir: Directory to save the output patches

Output Format

Patches are saved in Zarr format with the following structure:

• images: Array of image patches [N, C, H, W]
• labels: Array of label patches [N, H, W]
• positions: Array of patch locations [N, 2]
• metadata: Dictionary with additional information

A csv file is created of the zarr paths.

License

MIT License

Author

Victor Alhassan ([e-mail])

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.