Deep Active Re-Labeling: Toward Noise-Resilient Annotation Efficiency

Md Abdullah Al Forhad, Weishi Shi

While Deep Active Learning (DAL) effectively reduces human annotation costs, its efficacy is constrained by human annotation errors. This is because the data sampled for active lea…