P2-ML-GCN and RP2-ML-GCN.pytorch

Requirements

Please, install the following packages

• numpy
• torch-0.3.1
• torchnet
• torchvision-0.2.0
• tqdm

Please unzip the files data.zip and checkpoint.zip.

Download pretrain models

checkpoint/coco (GoogleDrive)

checkpoint/voc (GoogleDrive)

Laplace Noise

noise = torch.from_numpy(np.random.laplace(0,1/epsilon,feature.size())).cuda().float()

You can add the noise Laplace noise into the model's input features, weights and output (in P2_models.py).

Regularization term

The regularization term can be added into the original loss function (in RP2_demo_voc2007_gcn.py or RP2_demo_voc2007_gcn.py).

Train Voc2007

python3 RP2_demo_voc2007_gcn.py data/voc --image-size 448 --batch-size 16 --epochs 40

Train MS-COCO

python3 RP2_demo_coco_gcn.py data/coco --image-size 448 --batch-size 8 --epochs 20

Test Voc2007

python3 RP2_demo_voc2007_gcn.py data/voc --image-size 448 --batch-size 16 -e --resume checkpoint/voc/voc_checkpoint.pth.tar

Test MS-COCO

python3 RP2_demo_coco_gcn.py data/coco --image-size 448 --batch-size 8 -e --resume checkpoint/coco/coco_checkpoint.pth.tar

Training Framework

The framework follows ML-GCN Multi-Label Image Recognition with Graph Convolutional Networks, CVPR 2019. And add the Laplace noise into the dot-produce of features and weights.

Reference

This project is based on https://github.com/Megvii-Nanjing/ML-GCN

Tips

If you have any questions about our work, please do not hesitate to contact us.