This repository contains data and code for the paper GPS-Net: Graph Property Sensing Network for Scene Graph Generation (CVPR 2020).
If the paper significantly inspires you, we request that you cite our work:
@inproceedings{lin2020gps,
title={GPS-Net: Graph Property Sensing Network for Scene Graph Generation},
author={Lin, Xin and Ding, Changxing and Zeng, Jinquan and Tao, Dacheng},
booktitle={Proc. IEEE Conf. Comput. Vis. Pattern Recognit.},
pages={3746--3753},
year={2020}
}
Our reimplementations are based on rowanz/neural-motifs, and if you get any problem that cause you unable to run the project, you can check the issues under rowanz/neural-motifs first.
The Completed code is coming soon. But we have released some codes of DMP Modules, NPS-Loss and Frequency Softening. And you can see the codes in lib. Also, you can find more details in the draft version or our another implementation with a better object detection backbone.
Scene Graph Generation for Cityscape Video
Scene Graph Generation for Fruit Video
Scene Graph Generation for CCTV Video
| R@20 | R@50 | R@100 | |
|---|---|---|---|
| SGGen | 22.6 | 28.4 | 31.7 |
| SGCls | 36.1 | 39.2 | 40.1 |
| PredCls | 60.7 | 66.9 | 68.8 |