Abstract: Training images with data transformations, e.g., crops, shifts, rotations and color distortions, have been suggested as contrastive examples to evaluate the robustness of deep neural ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback