Search papers, labs, and topics across Lattice.
1
0
3
Forget GANs, this new method uses stylized fractals to generate synthetic training data that closes the reality gap and boosts ImageNet-100 accuracy by 10% when pre-training ViT-S.
