https://github.com/likesiwell/LGM-Net

Tensorflow code for ICML 2019 paper: LGM-Net: Learning to Generate Matching Networks for Few-Shot Learning
https://github.com/likesiwell/LGM-Net

Last synced: 13 days ago
JSON representation

Tensorflow code for ICML 2019 paper: LGM-Net: Learning to Generate Matching Networks for Few-Shot Learning

• Host: GitHub
• URL: https://github.com/likesiwell/LGM-Net
• Owner: likesiwell
• License: mit
• Created: 2019-05-11T05:22:55.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2020-04-27T15:07:04.000Z (over 4 years ago)
• Last Synced: 2024-04-10T06:07:07.685Z (7 months ago)
• Language: Python
• Size: 229 KB
• Stars: 83
• Watchers: 4
• Forks: 24
• Open Issues: 3
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-few-shot-meta-learning - code - official (TF)

README

        
# LGM-Net

TensorFlow source code for the following publication:

> LGM-Net: Learning to Generate Matching Networks for Few-Shot Learning

>

> Huaiyu Li, Weiming Dong, Xing Mei, Chongyang Ma, Feiyue Huang, Bao-Gang Hu

>

> In *Proceedings of the 36th International Conference on Machine Learning (ICML 2019)*

# Requirements

- Python 3.5

- [NumPy](http://www.numpy.org/)

- [SciPy](https://www.scipy.org/)

- [tqdm](https://pypi.python.org/pypi/tqdm)

- [Tensorflow 1.4](https://www.tensorflow.org/install/)

- [Opencv 3.2.0](https://opencv.org/)

## Preparation

Set the path of resized miniImageNet dataset in `data.py`

## Train

```

python train_meta_matching_network.py --way 5 --shot 1

python train_meta_matching_network.py --way 5 --shot 5

```

## Test

```

python train_meta_matching_network.py --way 5 --shot 1 --is_test True --ckp checkpoint_id

```

## Acknowledgements

Thanks to [Antreas Antoniou](https://github.com/AntreasAntoniou/) for his [Matching Networks implementation](https://github.com/AntreasAntoniou/MatchingNetworks) of which parts were used for this implementation.