This project hosts the code for implementing the ADD-GCN algorithm for multi-label image recognition, as presented in our paper:
Attention-Driven Dynamic Graph Convolutional Network for Multi-Label Image Recognition;
Jin Ye, Junjun He, Xiaojiang Peng, Wenhao Wu, Yu Qiao;
In: European Conference on Computer Vision (ECCV), 2020.
arXiv preprint arXiv:2012.02994
The full paper is available at: https://arxiv.org/abs/2012.02994.
After you have installed Pytorch, you can follow the below steps to run a quick demo.
python main.py --data COCO2014 --data_root_dir {YOUR-ROOT-DATA-DIR} --model_name ADD_GCN --resume {THE-TEST-MODEL} -e -i 448
Please note that:
-
You should put the COCO2014 folder in {YOUR-ROOT-DATA-DIR}.
-
You should put the test model in {THE-TEST-MODEL} folder.
-
You can get the same ADD-GCN results with this model. The password is
4ebj.
| Model | Test size | mAP |
|---|---|---|
| ResNet-101 | 448×448 | 79.7 |
| DecoupleNet | 448×448 | 82.2 |
| ML-GCN | 448×448 | 83.0 |
| ADD-GCN | 448×448 | 84.2 |
| ResNet-101 | 576×576 | 80.0 |
| SSGRL | 576×576 | 84.2 |
| ML-GCN | 576×576 | 84.3 |
| ADD-GCN | 576×576 | 85.2 |
Please consider citing our paper in your publications if the project helps your research. BibTeX reference is as follows.
@inproceedings{ye2020add,
title = {Attention-Driven Dynamic Graph Convolutional Network for Multi-Label Image Recognition},
author = {Jin Ye, Junjun He, Xiaojiang Peng, Wenhao Wu, Yu Qiao},
booktitle = {European Conference on Computer Vision (ECCV)},
year = {2020}
}