A Pytorch Implementation of Adaptive Image-to-video Scene Graph Generation via Knowledge Reasoning and Adversarial Learning (AAAI 2022)

Introduction

Follow faster-rcnn repository to setup the environment. We used Pytorch 0.4.1 and Python 3.6 for this project.

Download related data of VRD->VidVRD here.

An example of VRD->VidVRD

Train and test the detection model

bash ./scripts/instance_styleD_resnet101.sh

Train and test the SGG model

bash ./scripts/SGG_emb_resnet.sh

Citation

@article{chen2022image,
  title={Adaptive Image-to-video Scene Graph Generation via Knowledge Reasoning and Adversarial Learning},
  author={Jin Chen, Xiaofeng Ji, Xinxiao Wu,
  booktitle={The Thirty-Sixth {AAAI} Conference on Artificial Intelligence, {AAAI} 2022},
  year={2022},
}