Detect and Correct: A Selective Noise Correction Method for Learning with Noisy Labels

Yuval Grinberg, Nimrod Harel, Jacob Goldberger +1

Falsely annotated samples, also known as noisy labels, can significantly harm the performance of deep learning models. Two main approaches for learning with noisy labels are global…