Abstract: Image encoder pre-training has experienced a substantial evolution due to contrastive learning, facilitating the extraction of intricate feature representations from unlabeled datasets.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results