Skip to content

Latest commit

 

History

History
92 lines (79 loc) · 4.16 KB

README.md

File metadata and controls

92 lines (79 loc) · 4.16 KB

Cross-Domain Few-Shot Classification via Learned Feature-Wise Transformation

[Project Page][Paper]

Pytorch implementation for our cross-domain few-shot classification method. With the proposed learned feature-wise transformation layers, we are able to:

  1. improve the performance of exisiting few-shot classification methods under cross-domain setting
  2. achieve stat-of-the-art performance under single-domain setting.

Contact: Hung-Yu Tseng ([email protected])

Paper

Please cite our paper if you find the code or dataset useful for your research.

Cross-Domain Few-Shot Classification via Learned Feature-Wise Transformation
Hung-Yu Tseng, Hsin-Ying Lee, Jia-Bin Huang, Ming-Hsuan Yang
International Conference on Learning Representations (ICLR), 2020 (spotlight)

@inproceedings{crossdomainfewshot,
  author = {Tseng, Hung-Yu and Lee, Hsin-Ying and Huang, Jia-Bin and Yang, Ming-Hsuan},
  booktitle = {International Conference on Learning Representations},
  title = {Cross-Domain Few-Shot Classification via Learned Feature-Wise Transformation},
  year = {2020}
}

Usage

Prerequisites

  • Python >= 3.5
  • Pytorch >= 1.3 and torchvision (https://pytorch.org/)
  • You can use the requirements.txt file we provide to setup the environment via Anaconda.
conda create --name py36 python=3.6
conda install pytorch torchvision -c pytorch
pip3 install -r requirements.txt

Install

Clone this repository:

git clone https://github.com/hytseng0509/CrossDomainFewShot.git
cd CrossDomainFewShot

Datasets

Download 5 datasets seperately with the following commands.

  • Set DATASET_NAME to: cars, cub, miniImagenet, places, or plantae.
cd filelists
python3 process.py DATASET_NAME
cd ..
  • Refer to the instruction here for constructing your own dataset.

Feature encoder pre-training

We adopt baseline++ for MatchingNet, and baseline from CloserLookFewShot for other metric-based frameworks.

  • Download the pre-trained feature encoders.
cd output/checkpoints
python3 download_encoder.py
cd ../..
  • Or train your own pre-trained feature encoder (specify PRETRAIN to baseline++ or baseline).
python3 train_baseline.py --method PRETRAIN --dataset miniImagenet --name PRETRAIN --train_aug

Training with multiple seen domains

Baseline training w/o feature-wise transformations.

  • METHOD : metric-based framework matchingnet, relationnet_softmax, or gnnnet.
  • TESTSET: unseen domain cars, cub, places, or plantae.
python3 train_baseline.py --method METHOD --dataset multi --testset TESTSET --name multi_TESTSET_ori_METHOD --warmup PRETRAIN --train_aug

Training w/ learning-to-learned feature-wise transformations.

python3 train.py --method METHOD --dataset multi --testset TESTSET --name multi_TESTSET_lft_METHOD --warmup PRETRAIN --train_aug

Evaluation

Test the metric-based framework METHOD on the unseen domain TESTSET.

  • Specify the saved model you want to evaluate with --name (e.g., --name multi_TESTSET_lft_METHOD from the above example).
python3 test.py --method METHOD --name NAME --dataset TESTSET

Note

  • This code is built upon the implementation from CloserLookFewShot.
  • The dataset, model, and code are for non-commercial research purposes only.
  • You can change the number of shot (i.e. 1/5 shots) using the argument --n_shot.
  • You need a GPU with 16G memory for training the gnnnet approach w/ learning-to-learned feature-wise transformations.
  • 04/2020: We've corrected the code for training with multiple domains. Please find the link here for the model trained with the current implementation on Pytorch 1.4.