IO

class aeronet.dataset.io.Predictor(input_channels, output_labels, processing_fn, sample_size=(1024, 1024), bound=256, n_workers=1, verbose=True, **kwargs)

Bases: object

process(bc, output_directory)

class aeronet.dataset.io.SampleCollectionWindowWriter(directory, channels, shape, transform, crs, nodata, dtype='uint8')

Bases: object

close()

open()

write(raster, x, y, height, width, bounds=None)

class aeronet.dataset.io.SampleWindowWriter(fp, shape, transform, crs, nodata, dtype='uint8')

Bases: object

close()

height

open()

width

write(raster, x, y, width, height, bounds=None)

Writes the specified raster into a window in dst The raster boundaries can be cut by ‘bounds’ pixels to prevent boundary effects on the algorithm output. If width and height are not equal to size of raster (after the bounds are cut), which is not typical, the raster is stretched to the window size (width and height)

Parameters:

• raster – numpy array to be written into dst
• x – begin position of window
• y – begin position of window
• width – size of window
• height – size of window
• bounds – [[,][,]] - number of pixels to cut off from each side of the raster before writing

Returns:

class aeronet.dataset.io.SequentialSampler(band_collection, channels, sample_size, bound=0)

Bases: object