Can i exclude data augmentation for a particular class in multilabel dataset
talhaanwarch opened this issue · 0 comments
talhaanwarch commented
Is there a way to exclude or include a class from being augmented while dealing with multilabels.
Here is the flow from data frame code snipped
train_generator_1 = generator.flow_from_dataframe(dataframe,
directory=directory_path,
target_size=image_shape,
x_col="Image",
y_col=['C1' ,'C2' ,'C3', 'C4', 'C5' ,'C6' ,'C7', 'C8'],
class_mode='other',
shuffle=False,
batch_size=batch_size,
seed=7
)
The class C1, C2 and C3 have 10 time more instances than other classes, so i want to emphasis on minority classes. Is there a way to exclude C1, C2 and C3 from being augmented. Or minority classes to be augmented more than majority class.