This repository contains the code release of NEUCORE, from our paper:

NEUCORE: Neural Concept Reasoning for Composed Image Retrieval Shu Zhao, Huijuan Xu. UniReps Workshop @ Neural Information Processing Systems, 2023.

If this code and/or paper is useful in your research, please cite:

@article{zhao2023neucore,
  title={NEUCORE: Neural Concept Reasoning for Composed Image Retrieval},
  author={Zhao, Shu and Xu, Huijuan},
  journal={arXiv preprint arXiv:2310.01358},
  year={2023}
}

Create the environment for running our code as follow:

conda create --name neucore python=3.8
conda activate neucore
pip install -r requirements.txt

Note: using cuda version 11.1 (please modify requirements.txt otherwise)

Please refer to ARTEMIS repo to prepare datasets.

Datasets should be put into data folder.

Extract concepts from sentences by NLTK:

python scripts/extract_concepts_<dataset>.py

sh ./scripts/train.sh <dataset>

Pre-trained model: google drive

sh ./scripts/test.sh <dataset> <checkpoint_path>

Our code is based on ARTEMIS. Thanks for their great work!