Pytorch implementation for the CVPR 2021 paper: Towards Accurate Text-based Image Captioning with Content Diversity Exploration

Clone this repository, and build it with the following command.

# activate your own conda environment
# [Alternative]
# conda env create -f textcap.yaml
# conda activate textcap

git clone [email protected]:guanghuixu/AnchorCaptioner.git
cd AnchorCaptioner
python setup.py build develop

Some specific annotations required by our method are provided in here. More details please refer to projects/M4C_Captioner/README.md

If you use any part of our code in your research, please cite our paper:

@InProceedings{xu2021textcap,
  title = {Towards Accurate Text-based Image Captioning with Content Diversity Exploration},
  author = {Guanghui Xu and Mingkui Tan and Shuaicheng Niu and Yucheng Luo and Qing Du and Qi Wu},
  booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition},
  year = {2021}
}

The code is greatly inspired by the MMF and M4C-Captioner.