https://github.com/dingmyu/DAPN

A pytorch implementation of "Domain-Adaptive Few-Shot Learning"
https://github.com/dingmyu/DAPN

Last synced: about 2 months ago
JSON representation

A pytorch implementation of "Domain-Adaptive Few-Shot Learning"

• Host: GitHub
• URL: https://github.com/dingmyu/DAPN
• Owner: dingmyu
• Created: 2020-03-19T06:48:37.000Z (over 4 years ago)
• Default Branch: master
• Last Pushed: 2020-07-21T16:28:39.000Z (over 4 years ago)
• Last Synced: 2024-07-31T23:45:02.764Z (4 months ago)
• Language: Python
• Size: 288 KB
• Stars: 136
• Watchers: 2
• Forks: 26
• Open Issues: 6
• Metadata Files:
  • Readme: Readme.md

Awesome Lists containing this project

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

README

        
## Introduction

The framework is implemented and tested with Ubuntu 16.04, CUDA 8.0/9.0, Python 3, Pytorch 0.4/1.0/1.1, NVIDIA TITANX GPU. 

## Requirements

- **Cuda & Cudnn & Python & Pytorch**

    This project is tested with CUDA 8.0/9.0, Python 3, Pytorch 0.4/1.0, NVIDIA TITANX GPUs.

    Please install proper CUDA and CUDNN version, and then install Anaconda3 and Pytorch. Almost all the packages we use are covered by Anaconda.

- **My settings**

  ```shell

  source ~/anaconda3/bin/activate (python 3.6.5)

	(base)  pip list

	torch                              0.4.1

	torchvision                        0.2.2.post3

	numpy                              1.18.1

	numpydoc                           0.8.0

	numba                              0.42.0

	opencv-python                      4.0.0.21

  ```

## Data preparation

Download and unzip the datasets: **MiniImageNet, TieredImageNet, DomainNet**.

Here we provide the datasets of target domain in Google Drive, [miniImageNet](https://drive.google.com/file/d/1Yxzw2kJarXCV2tldKzXt6rlGqcLuv24W), [tieredImageNet](https://drive.google.com/file/d/1Unqwgiuoy7br8vKiEZo8Jhib-eNDxc5p).

Format:

(E.g. mini-imagenet)

  ```shell

MINI_DIR/

    --  train/

        --  n01532829/

        --  n01558993/

        ...

    --  train_new_domain/

    --  val/

    --  val_new_domain/

    --  test/

    --  test_new_domain/

  ```

## Training

First set the dataset path `MINI_DIR/, TIERED_DIR/, DOMAIN_DIR/` for the three datasets.

For each dataset, we use its training set to train a pre-trained model (e.g. tiered-imagenet).

``` 

cd pretrain/

python -u main_resnet.py --epochs 50 --batch_size 1024  --dir_path TIERED_DIR 2>&1 | tee log.txt &

```

We then use the corresponding pre-trained model to train on each dataset. (e.g. mini-imagenet)

```

python -u train_cross.py --gpu_id 0 --net ResNet50 --dset mini-imagenet --s_dset_path MINI_DIR --fsl_test_path MINI_DIR --shot 5 --train-way 16 --pretrained 'mini_checkpoint.pth.tar' --output_dir mini_way_16

```

## Testing

```

python -u test.py --load MODEL_PATH --root MINI_DIR

```