Learning with Noisy Labels

We propose a novel pseudo-label aware robust sample selection method for learning with noisy labels that outperforms state-of-the-art especially in presence of high label noise.