Awesome

Resnet18 Baseline

This repo contains the code for the following paper : Semantic Feature Augmentation in Few-shot Learning

We provided the code to reach our baseline performance in miniImagenet.(Resnet18+SVM)

We release the data split in split.rar.(CUB,caltech,cifar)

Datasets

Please put the data in:
/home/yourusername/data/miniImagenet

The images are put in 
.../miniImagenet/images
such as:miniImagenet\images\n0153282900000006.jpg
We provide the data split,please put them at 
.../miniImagenet/train.csv
.../miniImagenet/test.csv
.../miniImagenet/val.csv

Train

If you want to train a resnet18 from scratch by yourself:

python classification.py

You can also used our provided model

/samplecode/models/resnet18.t7

Then used it to do the one-shot task:

python SVM.py

Citation

@ARTICLE{semanticAugmentation, 
author={Z. {Chen} and Y. {Fu} and Y. {Zhang} and Y. {Jiang} and X. {Xue} and L. {Sigal}}, 
journal={IEEE Transactions on Image Processing}, 
title={Multi-level Semantic Feature Augmentation for One-shot Learning}, 
year={2019}, 
volume={}, 
number={}, 
pages={1-1}, 
keywords={one-shot learning;feature augmentation}, 
doi={10.1109/TIP.2019.2910052}, 
ISSN={1057-7149}, 
month={},
}