Notably, the performance gain of our method on the CUB200 dataset is not as significant as on the other two.
← all excerpts
Pursuing Better Representations: Balancing Discriminability and Transferability for Few-Shot Class-Incremental Learning.
1
—
—