The accuracy of the trained model depends on the accuracy of the ground truth, hence, spending the right amount of time and resources to ensure highly accurate data labelling is essential.