Remote Sensing Image Change Captioning With Dual-Branch Transformers: A New Method and a Large Scale Dataset

Chenyang Liu, Rui Zhao, Hao Chen, Zhengxia Zou, and Zhenwei Shi*✉

Share us a ⭐ if this repo does help

LEVIR-CC Dataset

Download Link

RSICCfromer

Here, we provide the pytorch implementation of the paper: "Remote Sensing Image Change Captioning With Dual-Branch Transformers: A New Method and a Large Scale Dataset".

For more information, please see our published paper in [IEEE | Lab Server] (Accepted by TGRS 2022)

Installation and Dependencies

git clone https://github.com/Chen-Yang-Liu/RSICC
cd RSICC
conda create -n RSICCformer_env python=3.6
conda activate RSICCformer_env
pip install -r requirements.txt

Data preparation

Firstly, put the downloaded dataset in ./LEVIR_CC_dataset/. Then preprocess dataset as follows:

python create_input_files.py --min_word_freq 5

After that, you can find some resulted files in ./data/.

Besides, the resulted files can also be downloaded from here: [Google Drive | Baidu Pan (code:nq9y)]. Extract it to ./data/.

!NOTE: For a fair comparison, we suggest that future researchers ensure min_word_freq <= 5 or use our preprocessed data above with min_word_freq = 5.

Inference Demo

You can download our RSICCformer pretrained model——by [Google Drive | Baidu Pan (code:2fbc)]

After downloaded the pretrained model, you can put it in ./models_checkpoint/.

Then, run a demo to get started as follows:

python caption.py --img_A ./Example/A/train_000016.png --img_B ./Example/B/train_000016.png --path ./models_checkpoint/

After that, you can find the generated caption in ./eval_results/

Train

Make sure you performed the data preparation above. Then, start training as follows:

python train.py  --data_folder ./data/ --savepath ./models_checkpoint/

Evaluate

python eval.py --data_folder ./data/ --path ./models_checkpoint/ --Split TEST

We recommend training 5 times to get an average score.

Citation:

@ARTICLE{9934924,
  author={Liu, Chenyang and Zhao, Rui and Chen, Hao and Zou, Zhengxia and Shi, Zhenwei},
  journal={IEEE Transactions on Geoscience and Remote Sensing}, 
  title={Remote Sensing Image Change Captioning With Dual-Branch Transformers: A New Method and a Large Scale Dataset}, 
  year={2022},
  volume={60},
  number={},
  pages={1-20},
  doi={10.1109/TGRS.2022.3218921}}

Reference:

Thanks to the following repository: a-PyTorch-Tutorial-to-Image-Captioning