Using Data Augmentation¶

Data augmentation performs random transformations on images given to a deep learning tool ^.NET ^Python during the training.

Experiment different settings to choose the best parameters for your data augmentation.
Configure data augmentation according to your problem. However, flips, shifts (20 - 40 px), brightness (5%), contrast (0.95 to 1.05) or salt and pepper noise (2%) can be useful on many datasets.
Check that the transformations do not change the label of an image (for example a defect that disappears because of a rotation or a contrast change).

Warning

With EasyLocate, we do not recommend to use rotation and shear data augmentation as it is not possible to compute the minimal bounding box surrounding the object after these geometric transformations.

Deep Learning Studio¶

Create and configure the data augmentation settings in the Data augmentation tab.
Display and review the data augmented images with the minimum settings (Lower limits augmentation), the maximum settings (Upper limits augmentation) or the random settings (Random augmentation).

Visualization of data augmentation settings.

API¶

Use an EDataAugmentation ^.NET ^Python object to configure the data augmentation settings. You can give this object to the to enable or disable these transformations or directly use an object EDataAugmentation ^.NET ^Python that you give to the Train method of your tool.

Available Transformations¶

Geometric Transformations¶

../../../_images/using_data_augmentation_2.webp — Geometric transformations¶

Horizontal and vertical flips (enabled with SetEnableHorizontalFlip ^.NET ^Python and SetEnableVerticalFlip ^.NET ^Python).
Scaling (between a minimum and maximum value defined with SetMinScale ^.NET ^Python and SetMaxScale ^.NET ^Python).
Horizontal and vertical shifts (between –maxValue and maxValue defined with SetMaxHorizontalShift ^.NET ^Python and SetMaxVerticalShift ^.NET ^Python).
Rotations (between 0 and a maximum value defined with SetMaxRotationAngle ^.NET ^Python).
Horizontal and vertical shear (between –maxValue and maxValue defined with SetMaxHorizontalShear ^.NET ^Python and SetMaxVerticalShear ^.NET ^Python).

Color and Luminosity Transformations¶

../../../_images/using_data_augmentation_3.webp — Color/Luminosity transformations¶

Brightness offset (between –maxValue and maxValue defined with SetMaxBrightnessOffset ^.NET ^Python).
Contrast gain (between a minimum and maximum value defined with SetMinContrastGain ^.NET ^Python and SetMaxContrastGain ^.NET ^Python).
Gamma corrections (between a minimum and maximum value defined with SetMinGamma ^.NET ^Python and SetMaxGamma ^.NET ^Python).
Hue offset (between –maxValue and maxValue defined with SetMaxHueOffset ^.NET ^Python).
Saturation gain (between a minimum and maximum value defined with SetMinSaturationGain ^.NET ^Python and SetMaxSaturationGain ^.NET ^Python).

Noise Transformations¶

../../../_images/using_data_augmentation_4.webp — Noise transformations¶

Gaussian noise, also called additive white noise, generated with a standard deviation (between a minimum and maximum value defined with SetGaussianNoiseMinimumStandardDeviation ^.NET ^Python and SetGaussianNoiseMaximumStandardDeviation ^.NET ^Python).
Speckle noise, a multiplicative noise, generated from a Gamma distribution with a mean of 1 and a standard deviation (between a minimum and a maximum value defined with SetSpeckleNoiseMinimumStandardDeviation ^.NET ^Python and SetSpeckleNoiseMinimumStandardDeviation ^.NET ^Python).
Salt and pepper noise generated from a pixel density (between a minimum and a maximum value defined with SetSaltAndPepperNoiseMinimumDensity ^.NET ^Python and SetSaltAndPepperNoiseMaximumDensity ^.NET ^Python).

Note

The standard deviation is expressed as a percentage of the maximum pixel value.