Sumeyyesaray·Oct 17DINO: The Evolution of Learning to See Without LabelsIn late 2020, a research team stumbled upon something strange. They trained a Vision Transformer with no labels at all, letting it learn…
Sumeyyesaray·Jul 23ViT vs Swin Transformer: Global Attention vs Windowed Attention — A Small Experiment on CIFAR-10In the first post of this series, ConvNeXt kept coming back to one comparison point: Swin Transformer. That was intentional on the ConvNeXt…
Sumeyyesaray·Jul 17ResNet-50 vs EfficientNet-B3 vs ConvNeXt-T: A Small Transfer Learning Experiment on CIFAR-10Most people pick a backbone the same way: whichever one shows up first in a tutorial, or whichever one “everyone” seems to be using. Few…