Computer Vision Toolbox Model for Vision Transformer Network

Implementation of several variants of the vision transformer (ViT) model.

The Vision Transformer (ViT) model is a pretrained transformer model for image classification. It is also used as a backbone for other computer vision tasks such as object detection. The support package consists of three variants of the ViT model:

• Base-16 model
• Small-16 model
• Tiny-16 model

Here, “base”, “small” and “tiny” represent the model architecture and size, and 16 represents the patch size hyper-parameter. Each variant has been pretrained on ImageNet data set with input resolution of 384 and is stored as a .MAT file.