Towards Calibrated Multi-Label Deep Neural Networks

Ours Figure 1. Left: Reliability diagram (calibration curve) of multi-label DNNs trained with the asymmetric focal loss [40], ASY loss [61] , and our proposed loss. Right: corresponding retrieval results on the multi-label retrieval task, where the user specifies a query string of desired labels P and undesired labels N . Correct retrieval results are highlighted in green. Improved calibration substantially improves retrieval performance.