Since the data is too large, we do not provide the data, including visual & textual scene graph, please generate the data by yourself using the code in data generation.
python>=3.6
pytorch
git submodule update --init --recursive
pip install spacy
python -m spacy download en
follow the repo's instruction to install Faster R-CNN
git clone https://github.com/shilrley6/Faster-R-CNN-with-model-pretrained-on-Visual-Genome.git
cd lib
python setup.py build develop
download pre-trained model
The code of visual scene graph generation is from Unbiased Scene Graph Generation from Biased Training
The code of textual scene graph generation is from SceneGraphParser
The code of image feature extraction is from Faster-R-CNN-with-model-pretrained-on-Visual-Genome