Simple method fixes classifier bias in imbalanced dataset learning

Original: Why Not Hyperparameter-Friendly Optimisation? A Monotonic Adaptive Norm Rescaling Approach For Long-Tailed Recognition

Source: arxiv.org ↗