Calibrating Deep Neural Networks using Focal Loss

Deep Neural Networks (DNNs) makes their predictions hard to rely on. Ideally, we want networks to be accurate, calibrated and confident.