🌍 MapMiner

MapMiner is a geospatial and model-centric tool designed to efficiently download, process, and analyze geospatial data and metadata from various sources. It leverages powerful Python libraries like Selenium, Dask, Numba, and Xarray to provide high-performance data handling and integrates state-of-the-art models for advanced geospatial AI and visualization.

🛠 Installation

Base installation:

pip install mapminer

Full installation (includes OCR + Chrome support):

pip install "mapminer[all]"

🚀 Key Features

• 🌐 Selenium: Automated web interactions for metadata extraction.
• ⚙️ Dask: Distributed computing to manage large datasets.
• 🚀 Numba: JIT compilation for accelerating numerical computations.
• 📊 Xarray: Multi-dimensional array data handling for seamless integration.

📚 Supported Datasets

MapMiner supports a variety of geospatial datasets across multiple categories:

Category	Datasets
🌍 Satellite	Sentinel-2, Sentinel-1, MODIS, Landsat
🚁 Aerial	NAIP
🗺️ Basemap	Google, ESRI
📍 Vectors	Google Building Footprint, OSM
🏔️ DEM (Digital Elevation Model)	Copernicus DEM 30m, ALOS DEM
🌍 LULC (Land Use Land Cover)	ESRI LULC
🌾 Crop Layer	CDL Crop Mask
🕒 Real-Time	Google Maps Real-Time Traffic

🧠 Supported Models

MapMiner provides pre-integrated state-of-the-art vision models for geospatial AI:

Model	Use Cases
🔥 DINOv3	Feature extraction, classification, segmentation, detection backbones
🌀 NAFNet	Denoising, deblurring, super-resolution, temporal consistency
⏳ ConvLSTM	Crop forecasting, temporal fusion (Sentinel-1/2), sequence modeling
💎 SAM3	Prompt-based instance segmentation, fast zero-shot object extraction

🤖 Models

1️⃣ DINOv3 Model

You can import DINOv3 directly for feature extraction or downstream tasks:

from mapminer.models import DINOv3
model = DINOv3(pretrained=True)
x = normalize(input_tensor)
output = model(x)

2️⃣ NAFNet Model

Use NAFNet for denoising, enhancement, or temporal SR tasks:

from mapminer.models import NAFNet
model = NAFNet(in_channels=12, dim=32)
output = model(input_tensor)

3️⃣ SAM3 Model

Use SAM3 for prompt-based instance segmentation on high-resolution geospatial imagery:

from mapminer.models import SAM3
sam3 = SAM3() 
df = sam3.inference(ds,text='building', exemplars=None)

⛏️ Miners

1️⃣ GoogleBaseMapMiner

from mapminer.miners import GoogleBaseMapMiner
miner = GoogleBaseMapMiner()
ds = miner.fetch(lat=40.748817, lon=-73.985428, radius=500)

2️⃣ CDLMiner

from mapminer.miners import CDLMiner
miner = CDLMiner()
ds = miner.fetch(lon=-95.665, lat=39.8283, radius=10000, daterange="2024-01-01/2024-01-10")

3️⃣ GoogleBuildingMiner

from mapminer.miners import GoogleBuildingMiner
miner = GoogleBuildingMiner()
ds = miner.fetch(lat=34.052235, lon=-118.243683, radius=1000)

🖼 Visualizing the Data

You can easily visualize the data fetched using hvplot:

import hvplot.xarray
ds.hvplot.image(title=f"Captured on {ds.attrs['metadata']['date']['value']}")

📦 Dependencies

MapMiner relies on several Python libraries:

• Selenium: For automated browser control.
• Dask: For distributed computing and handling large data.
• Numba: For accelerating numerical operations.
• Xarray: For handling multi-dimensional array data.
• EasyOCR: For extracting text from images.
• HvPlot: For visualizing xarray data.

🛠 Contributing

Contributions are welcome! Fork the repository and submit pull requests. Include tests for any new features or bug fixes.