Translation: The latest research indicates that Convolutional Neural Networks (ConvNets) can rival Vision Transformers on large-scale datasets, challenging previous notions. The research team trained the NFNet model on the enormous JFT-4B dataset and found that computational resources and the amount of available data are the primary influencers of model performance. The NFNet architecture demonstrates competitive capabilities on large-scale datasets. This study offers new insights for the field of computer vision.