COCA: Classifier-Oriented Calibration via Textual Prototype for Source-Free Universal Domain Adaptation.

Xinghong Liu, Yi Zhou

Browse the full ACCV paper archive.