Search papers, labs, and topics across Lattice.
3
0
5
3
InstructMixup achieves superior performance in visual model training by creating robust, label-consistent samples from a single image, outperforming nine competing methods.
Label-preserving self-saliency mixup can significantly boost model performance while maintaining semantic integrity in data augmentation.
Retraining just the classifier head of a frozen feature extractor can be dramatically improved by meta-learning feature-space augmentations that target hard examples, leading to state-of-the-art robustness against spurious correlations.
