Open SebastianBravo opened 1 year ago
I was looking at the GaussianNoise augmentation and you are not taking into account not to return negative values nor values greater than 1. Negative values on images might alter the training of models.
You can try using np.clip()
I was looking at the GaussianNoise augmentation and you are not taking into account not to return negative values nor values greater than 1. Negative values on images might alter the training of models.