This repository contains the official PyTorch implementation for the paper Toward Open-set Human Object Interaction Detection (AAAI2024).

We provide weights for DHD models trained on HICO-DET.

1. Install the lightweight deep learning library Pocket. The recommended PyTorch version is 1.9.0. Make sure the environment for Pocket is activated (conda activate pocket), and install the packaging library with pip install packaging.

2. init GroundingDINO and CLIP(from VIPLO).

3. Prepare the HICO-DET dataset.

   1. If you have not downloaded the dataset before, run the following script.

   cd /path/to/dhd/hicodet
   bash download.sh

   2. If you have previously downloaded the dataset, simply create a soft link.

   cd /path/to/dhd/hicodet
   ln -s /path/to/hicodet_20160224_det ./hico_20160224_det

4. Prepare the V-COCO dataset (contained in MS COCO).

   1. If you have not downloaded the dataset before, run the following script

   cd /path/to/dhd/vcoco
   bash download.sh

   2. If you have previously downloaded the dataset, simply create a soft link

   cd /path/to/dhd/vcoco
   ln -s /path/to/coco ./mscoco2014

5. Prepare the VG dataset from VG.

   1. If you have downloaded the dataset, simply create a soft link

   cd /path/to/dhd/vg
   ln -s /path/to/vg ./vg

6. Prepare the preprocessed annotations, from vg and hicodet, and put them into the corresponding dataset directory.

DHD is released under the BSD-3-Clause License.

Refer to train.sh for training and test.sh for testing commands with different options.

@inproceedings{wu2024toward,
  title={Toward Open-Set Human Object Interaction Detection},
  author={Wu, Mingrui and Liu, Yuqi and Ji, Jiayi and Sun, Xiaoshuai and Ji, Rongrong},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={38},
  number={6},
  pages={6066--6073},
  year={2024}
}

This repo is based on UPT.