RECALL achieves rehearsal-free continual learning for object classification by logit recall before new training, regression regularization, Mahalanobis loss on known categories, and new heads per sequence, outperforming prior methods on CORe50, iCIFAR-100, and the introduced HOWS-CL-25 dataset.
Rectified linear units improve restricted boltzmann machines,
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.CV 2verdicts
UNVERDICTED 2representative citing papers
The paper proposes an unsupervised domain alignment method using GANs with cycle consistency, adversarial, and SSIM losses to augment training data and reduce low-level dataset biases in computer vision.
citing papers explorer
-
RECALL: Rehearsal-free Continual Learning for Object Classification
RECALL achieves rehearsal-free continual learning for object classification by logit recall before new training, regression regularization, Mahalanobis loss on known categories, and new heads per sequence, outperforming prior methods on CORe50, iCIFAR-100, and the introduced HOWS-CL-25 dataset.
-
Unsupervised Domain Alignment to Mitigate Low Level Dataset Biases
The paper proposes an unsupervised domain alignment method using GANs with cycle consistency, adversarial, and SSIM losses to augment training data and reduce low-level dataset biases in computer vision.