vision-transformer-from-scratch

This repository includes several kinds of vision transformers from scratch so that one beginner can understand the theory of vision transformer easily. The basic transformer,the linformer transformer and the swin transformer are all trained and tested.

Requirements: PyTorch (>= 1.6.0); Python 3.6.9; Numpy (1.18.2); OpenCV ; Linformer;

Train the model: python main_train.py; In the main_train.py the basic transformer and the linformer can be selected.

Test the model: python test.py; In the main_train.py the basic transformer and the linformer can be selected.

The theory of vision transformer can reference the following document: https://towardsdatascience.com/implementing-visualttransformer-in-pytorch-184f9f16f632; https://www.kaggle.com/hannes82/vision-transformer-trained-from-scratch-pytorch;