Space Data
Intelligence Hub

Telemetry & Mission Formats

Format Comparison Chart

Reading GeoTIFF with Rasterio

import rasterio
import numpy as np
import matplotlib.pyplot as plt

# Open a Sentinel-2 Band 8 (NIR) GeoTIFF
with rasterio.open('sentinel2_B08.tif') as src:
    nir = src.read(1).astype('float32')
    crs = src.crs
    transform = src.transform
    meta = src.meta

# Open Red band (B04) for NDVI
with rasterio.open('sentinel2_B04.tif') as src:
    red = src.read(1).astype('float32')

# Calculate NDVI
ndvi = (nir - red) / (nir + red + 1e-10)

plt.figure(figsize=(10, 8))
plt.imshow(ndvi, cmap='RdYlGn', vmin=-1, vmax=1)
plt.colorbar(label='NDVI')
plt.title('Vegetation Index (NDVI)')
plt.savefig('ndvi_output.png', dpi=150)
plt.show()

Reading NetCDF with Xarray

import xarray as xr
import matplotlib.pyplot as plt
import cartopy.crs as ccrs

# Open ERA5 temperature NetCDF
ds = xr.open_dataset('era5_temperature_2023.nc')

# Select surface temperature for a specific time
temp = ds['t2m'].sel(time='2023-06-01T12:00')

# Convert Kelvin to Celsius
temp_c = temp - 273.15

# Plot with cartopy projection
fig, ax = plt.subplots(
    subplot_kw={'projection': ccrs.PlateCarree()},
    figsize=(14, 7)
)
temp_c.plot(ax=ax, transform=ccrs.PlateCarree(),
            cmap='RdBu_r', cbar_kwargs={'label':'°C'})
ax.coastlines()
ax.add_feature(cartopy.feature.BORDERS)
plt.title('ERA5 2m Temperature – June 2023')
plt.savefig('era5_temp.png')

GeoPandas Vector Analysis

import geopandas as gpd
from shapely.geometry import Point, Polygon
import rasterio.mask

# Load district boundaries shapefile
districts = gpd.read_file('india_districts.shp')
districts = districts.to_crs(epsg=4326)  # WGS-84

# Filter Telangana state
telangana = districts[
    districts['STATE_NAME'] == 'Telangana'
]

# Buffer 10 km around Hyderabad
hyd = telangana[telangana['DIST_NAME']=='Hyderabad']
hyd_proj = hyd.to_crs(epsg=32644)  # UTM for distance
hyd_buffer = hyd_proj.buffer(10000)  # 10 km

# Spatial join — find all parcels in buffer
parcels = gpd.read_file('parcels.shp').to_crs(epsg=32644)
nearby = gpd.sjoin(parcels, gpd.GeoDataFrame(
    geometry=hyd_buffer), how='inner')
print(f'Parcels within 10 km: {len(nearby)}')

PyTorch CNN for Image Classification

import torch
import torchvision.transforms as T
from torchvision.models import resnet18
from torch.utils.data import DataLoader
from torchvision.datasets import ImageFolder

# Transforms for satellite patch dataset
transform = T.Compose([
    T.Resize((64, 64)),
    T.ToTensor(),
    T.Normalize(mean=[0.485, 0.456, 0.406],
                std=[0.229, 0.224, 0.225])
])

# EuroSAT dataset — 27,000 Sentinel-2 patches
train_ds = ImageFolder('eurosat/train', transform)
loader   = DataLoader(train_ds, batch_size=32)

# Pretrained ResNet-18 fine-tuned for 10 classes
model = resnet18(pretrained=True)
model.fc = torch.nn.Linear(512, 10)
optimizer = torch.optim.Adam(model.parameters(), lr=1e-4)
criterion = torch.nn.CrossEntropyLoss()

model.train()
for imgs, labels in loader:
    optimizer.zero_grad()
    loss = criterion(model(imgs), labels)
    loss.backward()
    optimizer.step()