Attentive Graph Neural Networks for Few-Shot Learning

This repository contains the code for Attentive Graph Neural Networks for Few-Shot Learning.

Running the code

Preliminaries

Environment

Python 3.7.3
Pytorch 1.7.1
tensorboardX

Datasets

miniImageNet (courtesy of Spyros Gidaris)
tieredImageNet (courtesy of Kwonjoon Lee)

Download the datasets and link the folders into materials/ with names mini-imagenet, tiered-imagenet and imagenet. Note imagenet refers to ILSVRC-2012 1K dataset with two directories train and val with class folders.

When running python programs, use --gpu to specify the GPUs for running the code (e.g. --gpu 0,1). For Classifier-Baseline, we train with 4 GPUs on miniImageNet and tieredImageNet and with 8 GPUs on ImageNet-800. Meta-Baseline uses half of the GPUs correspondingly.

In following we take miniImageNet as an example. For other datasets, replace mini with tiered. By default it is 1-shot, modify shot in config file for other shots. Models are saved in save/.

*The models on miniImageNet and tieredImageNet use ConvNet-4 as backbone, the channels in each block are 64-96-128-256.

1. Pretraining the ConvNet4 Backbone

python train_classifier.py --config configs/train_classifier_mini.yaml

The pretrained Classifier-Baselines can be downloaded from Google Drive (Mini-ImageNet , Tiered-ImageNet).

You can unzip and place the foder under the 'save' folder.

2. Training and Testing AGNN

python train_meta.py --config configs/train_meta_mini.yaml

Citation

@inproceedings{cheng2022attentive,
  title={Attentive graph neural networks for few-shot learning},
  author={Cheng, Hao and Zhou, Joey Tianyi and Tay, Wee Peng and Wen, Bihan},
  booktitle={2022 IEEE 5th International Conference on Multimedia Information Processing and Retrieval (MIPR)},
  pages={152--157},
  year={2022},
  organization={IEEE}
}

Extension is accepted by IEEE Transactions on Multimedia, Update Soon

Acknowledgment

We thank the following repos providing helpful components/functions in our work.

wenbihan / AGNN