in the first ungraded lab : Ungraded Lab Data Augmentation
the valid accuracy befor augmentation was 0.7320 and becomes after augmentation : 0.7270
data augmentation hurted the validation accuracy ! so why we apply it ?
in the first ungraded lab : Ungraded Lab Data Augmentation
the valid accuracy befor augmentation was 0.7320 and becomes after augmentation : 0.7270
data augmentation hurted the validation accuracy ! so why we apply it ?
That’s not a statistically significant difference.
I think we need more epoch with augmentation to reach bigger accuracy without overfitting