/SwinFG

This is the implementation of paper: SwinFG: A fine-grained recognition scheme based on swin transformer

Primary LanguagePython

SwinFG: A Fine-grained Recognition Scheme Based on Swin Transformer

This is the implementation of paper: SwinFG: A fine-grained recognition scheme based on swin transformer

Framework of SwinFG

Dependencies:

  • Python 3.8.12
  • PyTorch 1.5.1
  • torchvision 0.6.1

Usage

1. Pre-trained model

You can download the official pre-trained model file of the swin Transformer from here.

2. Prepare data

The dataset we used can be downloaded here:

3. Install required packages

pip3 install -r requirements.txt

4. Train

CUDA_VISIBLE_DEVICES=0  python3 -m torch.distributed.launch --nproc_per_node 1 train.py --dataset CUB --name myTrain

4. Citing SwinFG

@article{MA2024123021,
title = {SwinFG: A fine-grained recognition scheme based on swin transformer},
journal = {Expert Systems with Applications},
volume = {244},
pages = {123021},
year = {2024},
issn = {0957-4174},
doi = {https://doi.org/10.1016/j.eswa.2023.123021},
author = {Zhipeng Ma and Xiaoyu Wu and Anzhuo Chu and Lei Huang and Zhiqiang Wei}
}