Large-Scale Long-Tailed Recognition in an OpenWorld

Open Long-Tailed Recognition(OLTR):從自然分佈的數據中學習並在包括頭、尾和開放類的平衡測試集上優化分類精度。也就是對三個問題的綜合 imbalanced classification few-shot learning open-set recognition. 論文針對的問題： OLTR不僅要在封閉環境中處理不平衡分類和few-shot learning問題，還

>>阅读原文<<