Abstract: We investigate methods for combining multiple self-supervised tasks-i.e., supervised tasks where data can be collected without manual labeling-in order to train a single visual ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback