Rasterize ↔ vectorize¶

Cross the raster/vector divide both ways:

• Vectorize a raster: to_feature_collection() (polygonize equal-valued regions) and contour(interval=...) (iso-value lines).
• Rasterize a vector: Dataset.from_features(fc, ...) burns an attribute column into a new raster grid.

Setup¶

In [1]:

%matplotlib inline

import tempfile
from pathlib import Path

import numpy as np

DATA = Path('../../../examples/data')
WORK = Path(tempfile.mkdtemp(prefix='pyramids-ops-'))
DATA.is_dir(), WORK.is_dir()

%matplotlib inline import tempfile from pathlib import Path import numpy as np DATA = Path('../../../examples/data') WORK = Path(tempfile.mkdtemp(prefix='pyramids-ops-')) DATA.is_dir(), WORK.is_dir()

Out[1]:

(True, True)

In [2]:

from pyramids.dataset import Dataset
from pyramids.feature import FeatureCollection

ds = Dataset.read_file(DATA / 'acc4000.tif')
ds.shape, ds.epsg

from pyramids.dataset import Dataset from pyramids.feature import FeatureCollection ds = Dataset.read_file(DATA / 'acc4000.tif') ds.shape, ds.epsg

2026-07-11 14:42:10 | INFO | pyramids.base.config | Logging is configured.

Out[2]:

((1, 13, 14), 32618)

Vectorize — polygonize¶

to_feature_collection(add_geometry='Polygon') merges equal-valued cells into polygons.

In [3]:

polys = ds.to_feature_collection(add_geometry='Polygon')
type(polys).__name__, len(polys), list(polys.columns)

polys = ds.to_feature_collection(add_geometry='Polygon') type(polys).__name__, len(polys), list(polys.columns)

Out[3]:

('GeoDataFrame', 89, ['Band_1', 'geometry'])

Plot the polygons produced by polygonizing the raster — each equal-valued region becomes a polygon.

In [4]:

polys.plot(column='Band_1')

polys.plot(column='Band_1')

Out[4]:

<Axes: >

Vectorize — contour lines¶

contour(interval=...) traces iso-value lines into a FeatureCollection.

In [5]:

lines = ds.contour(interval=10.0)
type(lines).__name__, len(lines)

lines = ds.contour(interval=10.0) type(lines).__name__, len(lines)

Out[5]:

('FeatureCollection', 19)

Rasterize — burn a vector into a raster¶

Dataset.from_features rasterises an attribute column onto a new grid. Give it a numeric column to burn and a cell_size (or a template raster to inherit the grid).

In [6]:

zones = FeatureCollection.read_file(DATA / 'coello_polygons.geojson')
zones['value'] = np.arange(1, len(zones) + 1, dtype='int32')  # numeric column to burn
burned = Dataset.from_features(zones, cell_size=ds.cell_size, column_name='value')
burned.shape, burned.epsg

zones = FeatureCollection.read_file(DATA / 'coello_polygons.geojson') zones['value'] = np.arange(1, len(zones) + 1, dtype='int32') # numeric column to burn burned = Dataset.from_features(zones, cell_size=ds.cell_size, column_name='value') burned.shape, burned.epsg

Out[6]:

((1, 13, 13), 32618)

The round-trip, visualized: the source vector zones (coloured by the value column) and the raster they were burned into.

In [7]:

zones.plot(column='value')
burned.plot(band=0, title='rasterized zones')

zones.plot(column='value') burned.plot(band=0, title='rasterized zones')

Out[7]:

<cleopatra.array_glyph.ArrayGlyph at 0x7fded0f5a3c0>

Notes¶

• to_feature_collection() without add_geometry returns a plain table of cell values; pass add_geometry='Polygon' (or 'Point') to attach geometry.
• from_features(..., template=ds) inherits an existing raster's grid instead of cell_size.
• sieve() removes speckle (small regions) before vectorizing — useful on classified rasters.
• See also: Zonal statistics, Crop & mask.