Knowledge Guided Simple Primitives

This is the PyTorch code our CVPR 2022 work KG-SP: Knowledge Guided Simple Primitives for Open World Compositional Zero-Shot Learning . The code provides the implementation of our proposed method KG-SP, along with the baselines of CompCos and CGE taken from here. Additionally, we also provide the splits for our pCZSL setting on 3 datasets (UT-Zappos, MIT-States and C-GQA).

Setup

1. Clone the repo

2. We recommend using Anaconda for environment setup. To create the environment and activate it, please run:

    conda env create --file environment.yml
    conda activate czsl

4. Go to the cloned repo and open a terminal. Download the datasets and embeddings, specifying the desired path (e.g. DATA_ROOT in the example):

    bash ./utils/download_data.sh DATA_ROOT
    mkdir logs

Training

Open World. To train a model, the command is simply:

    python train.py --config CONFIG_FILE 

where CONFIG_FILE is the path to the configuration file of the model. The folder configs contains configuration files for all methods, i.e. CGE in configs/cge, CompCos in configs/compcos, and the other methods in configs/baselines.

To run KG-SP on MIT-States, the command is just:

    python train.py --config configs/kgsp/mit.yml --open_world --fast

On UT-Zappos, the command is:

    python train.py --config configs/kgsp/utzappos.yml --open_world --fast

Partial Label Setting To train KG-SP (in the partial label setting) on MIT-States, run:

    python train.py --config configs/kgsp/partial/mit.yml --partial --fast

Note: To create a new config, all the available arguments are indicated in flags.py.

Test

Open World. To test a model in the open world setting, run:

    python test.py --logpath LOG_DIR --open_world --fast

Partial Label Setting To test a KG-SP model on the partial label setting, a similar command can be used:

    python test.py --logpath LOG_DIR --fast --partial

References

If you use this code, please cite

@inproceedings{karthik2022open,
  title={KG-SP: Knowledge Guided Simple Primitives for Open World Compositional Zero-Shot Learning},
  author={Karthik, S and Mancini, M and Akata, Zeynep},
  booktitle={35th IEEE Conference on Computer Vision and Pattern Recognition},
  year={2022},
  organization={IEEE}
}

and

@inproceedings{mancini2021open,
  title={Open World Compositional Zero-Shot Learning},
  author={Mancini, M and Naeem, MF and Xian, Y and Akata, Zeynep},
  booktitle={34th IEEE Conference on Computer Vision and Pattern Recognition},
  year={2021},
  organization={IEEE}
}