Skip to content

module deeplearning

class DL

Deep Learning classification in Remote Sensing

method __init__

1
__init__(image, endmembers, nodata=-99999)

Parameter:

image: Optical images. It must be rasterio.io.DatasetReader with 3d.

endmembers: Endmembers must be a matrix (numpy.ndarray) and with more than one endmember. Rows represent the endmembers and columns represent the spectral bands. The number of bands must be equal to the number of endmembers. E.g. an image with 6 bands, endmembers dimension should be $n*6$, where $n$ is rows with the number of endmembers and 6 is the number of bands (should be equal). In addition, Endmembers must have a field (type int or float) with the names of classes to be predicted.

nodata: The NoData value to replace with -99999.

method FullyConnected

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
FullyConnected(
    hidden_layers=3,
    hidden_units=[64, 32, 16],
    output_units=10,
    input_shape=(6,),
    epochs=300,
    batch_size=32,
    training_split=0.8,
    random_state=None
)

This algorithm consists of a network with a sequence of Dense layers, which area densely connnected (also called fully connected) neural layers. This is the simplest of deep learning.

Parameters:

  • hidden_layers: Number of hidden layers to be used. 3 is for default.

  • hidden_units: Number of units to be used. This is related to 'neurons' in each hidden layers.

  • output_units: Number of clases to be obtained.

  • input_shape: The input shape is generally the shape of the input data provided to the Keras model while training. The model cannot know the shape of the training data. The shape of other tensors(layers) is computed automatically.

  • epochs: Number of iteration, the network will compute the gradients of the weights with regard to the loss on the batch, and update the weights accordingly.

  • batch_size: This break the data into small batches. In deep learning, models do not process antire dataset at once.

  • training_split: For splitting samples into two subsets, i.e. training data and for testing data.

  • random_state: Random state ensures that the splits that you generate are reproducible.

  • Please, see for more details https: //scikit-learn.org/stable/modules/generated/sklearn.model_selection.train_test_split.html

Return:

A dictionary with Labels of classification as numpy object, overall accuracy, among others results.

This file was automatically generated via lazydocs.