# Masked image patch prediction

BEiT is a Vision Transformer-based model pre-trained on ImageNet-21k through self-supervised learning and fine-tuned on ImageNet-1k for image classification tasks.