Sangh0/Vision-Transformer

Implementation of a Paper related to Vision Transformer

Implement and Review paper related to Vision Transformer

Implementation or Review Paper List

• ViT (ICLR 2021) (classification or backbone network)
• DeiT (ICML 2021) (classification or backbone network with knowledge distillation)
• Swin Transformer (ICCV 2021) (classification or backbone network)
• Segmenter (ICCV 2021) (semantic segmentation)
• SETR (CVPR 2021) (semantic segmentation)
• CoaT (ICCV 2021) (classification or backbone network)

What is Vision Transformer?

• Vision Transformer is a model for image classification inspired by Transformer in NLP
• This deal with image patches like words in NLP
• Self-Attention is important main idea
• It dosen't use CNN completely and achieved SOTA in each task of Vision