Query-aware Cross-mixup and Cross-reconstruction for Few-shot Fine-grained Image Classification

Code environment

This code requires Pytorch 1.7.0 and torchvision 0.8.0 or higher with cuda support. It has been tested on Ubuntu 16.04.

You can create a conda environment with the correct dependencies using the following command lines:

conda env create -f environment.yml

Dataset

You must first specify the value of data_path in config.yml.

The following datasets are used in our paper:

• CUB_200_2011 [Dataset Page]
• FGVC-Aircraft [Dataset Page]
• Flowers [Dataset Page]
• Stanford-Cars [Paper Page]

The following folders will exist in your data_path:

• CUB_fewshot_cropped: 100/50/50 classes for train/validation/test, using bounding-box cropped images as input
• Aircraft_fewshot: 50/25/25 classes for train/validation/test
• Flowers: 51/26/25 classes for train/validation/test
• Stanford-Cars: 130/17/49 classes for train/validation/test

Train

For example, to train QCCN on CUB_fewshot_cropped with ResNet-12 as the network backbone, run the following command lines:

cd experiments/CUB_fewshot_cropped/QCCN/ResNet-12/
./train.sh

Test

For example, to test QCCN on CUB_fewshot_cropped with ResNet-12 as the network backbone under the 5-way 1-shot and 5-way 5-shot setting, run the following command lines:

cd experiments/CUB_fewshot_cropped/QCCN/ResNet-12/
python test.py