str's Introduction

STR

This is the code implementation of "Show, Tell and Rephrase: Diverse Video Captioning via Two-Stage Progressive Training" published in TMM.

Introduction

This code provides the code implementation of the methods described in the article. We propose a novel progressive training method by decomposing the learning of latent variables into two stages that are topic-oriented and paraphrase-oriented, respectively.

Code Execution

The following steps explain how to run the code:

Install Dependencies: Ensure that the necessary dependencies are installed on your system

python -m venv myenv (optional)
source myenv/bin/activate (optional)
pip install -r requirements.txt

Data Preparation: Prepare the dataset and place it in the specified path. (You can download from Tsinghua Cloud)
- datasets directory: MSVD, MSRVTT, VATEX etc.
- cococaption: cococaption
Configuration: Modify the parameters in the configuration file in the configs directory according to your needs.
Run the Code: Execute the command in bash.sh to run the code. (train or evaluate)

NOTE: We release the optimal checkpoints for MSVD and MSRVTT in this onedrive link. You can put them into the results directory.

After execution, the results will be saved in the specified output path (default is results/).

Citation

If you use the code in your research, please cite the following article:

@article{liu2022show,
  title={Show, Tell and Rephrase: Diverse Video Captioning via Two-Stage Progressive Training},
  author={Liu, Zhu and Wang, Teng and Zhang, Jinrui and Zheng, Feng and Jiang, Wenhao and Lu, Ke},
  journal={IEEE Transactions on Multimedia},
  year={2022},
  publisher={IEEE}
}