TrainImagesClassifier (ann)¶

Описание¶

<put algortithm description here>

Parameters¶

Input Image List [multipleinput: rasters]

<put parameter description here>

Input Vector Data List [multipleinput: any vectors]

<put parameter description here>

Input XML image statistics file [file]

Optional.

<put parameter description here>

Default elevation [number]

<put parameter description here>

Default: 0

Maximum training sample size per class [number]

<put parameter description here>

Default: 1000

Maximum validation sample size per class [number]

<put parameter description here>

Default: 1000

On edge pixel inclusion [boolean]

<put parameter description here>

Default: True

Training and validation sample ratio [number]

<put parameter description here>

Default: 0.5

Name of the discrimination field [string]

<put parameter description here>

Default: Class

Classifier to use for the training [selection]

<put parameter description here>

Options:

• 0 — ann

Default: 0

Train Method Type [selection]

<put parameter description here>

Options:

• 0 — reg
• 1 — back

Default: 0

Number of neurons in each intermediate layer [string]

<put parameter description here>

Default: None

Neuron activation function type [selection]

<put parameter description here>

Options:

• 0 — ident
• 1 — sig
• 2 — gau

Default: 1

Alpha parameter of the activation function [number]

<put parameter description here>

Default: 1

Beta parameter of the activation function [number]

<put parameter description here>

Default: 1

Strength of the weight gradient term in the BACKPROP method [number]

<put parameter description here>

Default: 0.1

Strength of the momentum term (the difference between weights on the 2 previous iterations) [number]

<put parameter description here>

Default: 0.1

Initial value Delta_0 of update-values Delta_{ij} in RPROP method [number]

<put parameter description here>

Default: 0.1

Update-values lower limit Delta_{min} in RPROP method [number]

<put parameter description here>

Default: 1e-07

Termination criteria [selection]

<put parameter description here>

Options:

• 0 — iter
• 1 — eps
• 2 — all

Default: 2

Epsilon value used in the Termination criteria [number]

<put parameter description here>

Default: 0.01

Maximum number of iterations used in the Termination criteria [number]

<put parameter description here>

Default: 1000

set user defined seed [number]

<put parameter description here>

Default: 0

Outputs¶

Output confusion matrix [file]

<put output description here>

Output model [file]

<put output description here>

Console usage¶

processing.runalg('otb:trainimagesclassifierann', -io.il, -io.vd, -io.imstat, -elev.default, -sample.mt, -sample.mv, -sample.edg, -sample.vtr, -sample.vfn, -classifier, -classifier.ann.t, -classifier.ann.sizes, -classifier.ann.f, -classifier.ann.a, -classifier.ann.b, -classifier.ann.bpdw, -classifier.ann.bpms, -classifier.ann.rdw, -classifier.ann.rdwm, -classifier.ann.term, -classifier.ann.eps, -classifier.ann.iter, -rand, -io.confmatout, -io.out)

See also¶