lfcnn.metrics package
Submodules
lfcnn.metrics.metrics module
- class lfcnn.metrics.metrics.BadPix01(*args, **kwargs)[source]
Bases:
MeanMetricWrapper
Calculate the amount of pixels in percent that deviate more than 0.01 from the true value.
- class lfcnn.metrics.metrics.BadPix03(*args, **kwargs)[source]
Bases:
MeanMetricWrapper
Calculate the amount of pixels in percent that deviate more than 0.03 from the true value.
- class lfcnn.metrics.metrics.BadPix07(*args, **kwargs)[source]
Bases:
MeanMetricWrapper
Calculate the amount of pixels in percent that deviate more than 0.07 from the true value.
- class lfcnn.metrics.metrics.CosineProximity(*args, **kwargs)[source]
Bases:
MeanMetricWrapper
Computes the cosine proximity of y_pred and y_true.
- class lfcnn.metrics.metrics.MeanAbsoluteError(*args, **kwargs)[source]
Bases:
MeanMetricWrapper
Computes the mean absolute error of y_pred and y_true.
- class lfcnn.metrics.metrics.MeanSquaredError(*args, **kwargs)[source]
Bases:
MeanMetricWrapper
Computes the mean squared error of y_pred and y_true.
- class lfcnn.metrics.metrics.MultiScaleStructuralSimilarity(*args, **kwargs)[source]
Bases:
MeanMetricWrapper
Computes the multiscale structural similarity of y_pred and y_true.
- class lfcnn.metrics.metrics.PSNR(*args, **kwargs)[source]
Bases:
MeanMetricWrapper
Computes the PSNR of y_pred and y_true.
- class lfcnn.metrics.metrics.SpectralInformationDivergence(*args, **kwargs)[source]
Bases:
MeanMetricWrapper
Computes the spectral information divergence of y_pred and y_true.
- class lfcnn.metrics.metrics.StructuralSimilarity(*args, **kwargs)[source]
Bases:
MeanMetricWrapper
Computes the structural similarity of y_pred and y_true.
- class lfcnn.metrics.metrics.TotalVariation(*args, **kwargs)[source]
Bases:
MeanMetricWrapper
Computes the total variation of y_pred.
- lfcnn.metrics.metrics.get_central_metrics_fullsize(k1=0.01, k2=0.03, ssim_filter_size=11, ms_ssim_filter_size=11, ms_ssim_power_factors=(0.0448, 0.2856, 0.3001, 0.2363, 0.1333))[source]
Returns metrics used to evaluate multispectral central views.
The SSIM and MS-SSIM values are set with the values presented in the according papers [1, 2]. In particular, the MS-SSIM is calculated on 5 scales with a filter length of 11. Hence, the input images should have a spatial resolution of at least 256 x 256.
- Returns:
MeanAbsoluteError, MeanSquaredError, PSNR StructuralSimilarity, MultiscaleStructuralSimilarity and NormalizedCosineProximity
- Return type:
List of Metric instances containing
- lfcnn.metrics.metrics.get_central_metrics_small()[source]
Returns metrics used to evaluate multispectral central views.
The SSIM and MS-SSIM values are adapted to be used with small sized images of spatial resolutions around 32 x 32. The SSIM filter size is chosen to be 5. The MS-SSIM filter size is chosen to be 3. The MS-SSIM is calculated on 3 scales with power factors 0.5, 0.3, 0.2
- Returns:
MeanAbsoluteError, MeanSquaredError, PSNR StructuralSimilarity, MultiscaleStructuralSimilarity and NormalizedCosineProximity
- Return type:
List of Metric instances containing
Module contents
The LFCNN metrics module.