Data augmentation Magnitudes

[MFlossmann]

Hello! In the paper (Section 4, Subsection D) Data augmentation), you mention Data Augmentation with randomly sampled magnitudes. Unfortunately, random magnitudes of the transformations introduces quite a lot of hyperparameters to tune when trying to reproduce the paper.

Could you please supply the extends of the magnitudes of the augmentations you used in the paper as a baseline for our reproduction work?

I'd be specifically interested in the Minima/Maxima of the distributions concerning:

• Contrast
• Brightness
• Salt and Pepper magnitude
• Gaussian Noise Sigma

Als well as the Kernel Size of the gaussian Blur.

Thank you in advance.

[Teslatic]

@felipecode I would also like to know these parameter.

[felipecode]

Here it goes:

using IMGAUG

    st = lambda aug: iaa.Sometimes(0.4, aug)
    oc = lambda aug: iaa.Sometimes(0.3, aug)
    rl = lambda aug: iaa.Sometimes(0.09, aug)
    self.augment = iaa.Sequential([

        rl(iaa.GaussianBlur((0, 1.5))),  # blur images with a sigma between 0 and 1.5
        rl(iaa.AdditiveGaussianNoise(loc=0, scale=(0.0, 0.05), per_channel=0.5)),  # add gaussian noise to images
        oc(iaa.Dropout((0.0, 0.10), per_channel=0.5)),  # randomly remove up to X% of the pixels
        oc(iaa.CoarseDropout((0.0, 0.10), size_percent=(0.08, 0.2), per_channel=0.5)),
        # randomly remove up to X% of the pixels
        oc(iaa.Add((-40, 40), per_channel=0.5)),  # change brightness of images (by -X to Y of original value)
        st(iaa.Multiply((0.10, 2.5), per_channel=0.2)),  # change brightness of images (X-Y% of original value)
        rl(iaa.ContrastNormalization((0.5, 1.5), per_channel=0.5)),  # improve or worsen the contrast
        rl(iaa.Grayscale((0.0, 1))),  # put grayscale

    ],
        random_order=True  # do all of the above in random order
    )

[pri2si17-1997]

Hello @felipecode

Can you please explain what is st, oc and rl? What I can infer is it is sometimes augmenter. But then why we are using it in three different ways?

[felipecode]

Hey @pri2si17-1997 It means the frequency the augmentation is applyed.

St means for 40% of the images ( sometimes) Oc is for 30% of the images ( ocasionally) Rl is for 9& of the images ( rare)

[Suryavf]

Here it goes:

using IMGAUG

    st = lambda aug: iaa.Sometimes(0.4, aug)
    oc = lambda aug: iaa.Sometimes(0.3, aug)
    rl = lambda aug: iaa.Sometimes(0.09, aug)
    self.augment = iaa.Sequential([

        rl(iaa.GaussianBlur((0, 1.5))),  # blur images with a sigma between 0 and 1.5
        rl(iaa.AdditiveGaussianNoise(loc=0, scale=(0.0, 0.05), per_channel=0.5)),  # add gaussian noise to images
        oc(iaa.Dropout((0.0, 0.10), per_channel=0.5)),  # randomly remove up to X% of the pixels
        oc(iaa.CoarseDropout((0.0, 0.10), size_percent=(0.08, 0.2), per_channel=0.5)),
        # randomly remove up to X% of the pixels
        oc(iaa.Add((-40, 40), per_channel=0.5)),  # change brightness of images (by -X to Y of original value)
        st(iaa.Multiply((0.10, 2.5), per_channel=0.2)),  # change brightness of images (X-Y% of original value)
        rl(iaa.ContrastNormalization((0.5, 1.5), per_channel=0.5)),  # improve or worsen the contrast
        rl(iaa.Grayscale((0.0, 1))),  # put grayscale

    ],
        random_order=True  # do all of the above in random order
    )

Does this code require to normalize input image?