Gary-code / Ob-VPG

Code for ACM Trans. Multim. Comput. Commun. Appl. (TOMM) 2023 paper "Visual Paraphrase Generation with Key Information Retained"

Ob-VPG: Object-level Visual Paraphrase Generation

Released code for paper Visual Paraphrase Generation with Key Information Retained in TOMM 2023.

Requirements and Setup

1. Install Anaconda or Miniconda distribution based on Python3+ from their downloads' site.
2. Install python package the code needs.

Dataset & Pretrain VisualBERT model

All the training, validation and test data in the data_sentences folder.

• All this data is preprocessed from the MSCOCO caption dataset.
• More details about preprocessing, you can see in repository

You can download visualBERT model from link

Training & Evaluation

python train.py

Reference

@article{xie2023visual,
  title={Visual paraphrase generation with key information retained},
  author={Xie, Jiayuan and Chen, Jiali and Cai, Yi and Huang, Qingbao and Li, Qing},
  journal={ACM Transactions on Multimedia Computing, Communications and Applications},
  volume={19},
  number={6},
  pages={1--19},
  year={2023},
  publisher={ACM New York, NY}
}