GradMix masks the most attribution-activated image regions during training, pushing the model to learn additional features and improving open set recognition and robustness.
We list the final applied hyper-parameters in Table 8, including γ andλ in the learning objective,k in the detection method, the number of training epochs, and batch size (BS)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Informed Mixing -- Improving Open Set Recognition via Attribution-based Augmentation
GradMix masks the most attribution-activated image regions during training, pushing the model to learn additional features and improving open set recognition and robustness.