ReMarNet: Conjoint Relation and Margin Learning for Small-Sample Image Classification

Code release for the paper ReMarNet: Conjoint Relation and Margin Learning for Small-Sample Image Classification (TCSVT 2020).

Dataset

UIUC-Sports

• UIUC-Sports contains 1578 sports scene images of 8 classes: bocce (137), polo (182),rowing (250), sailing (190), snowboarding (190), rock climb-ing (194), croquet (236) and badminton (200). A training set of 749 images and a test set of 749 images are randomly sampled from the entire dataset.
• You can download the dataset at http://vision.stanford.edu/lijiali/event_dataset/. (Li-Jia Li and Li Fei-Fei. What, where and who? Classifying event by scene and object recognition . IEEE Intern. Conf. in Computer Vision (ICCV). 2007 (PDF) )
• The style of our randomly divided dataset is shown in the related Excel table. You can divide the dataset according to the name of the sample in our table.

Requirements

• python=2.7
• PyTorch=1.4.0
• torchvision=0.5.0
• pillow=6.2.1
• numpy=1.15.4

Training

• Download datasets
• You can run the code using the following command:

python UIUC_ReMarNet.py
python UIUC_Baseline.py 

Results

Dataset	Measure	Baseline	Ours
UIUC-Sports	Mean	0.9476	0.9581
	Std.	0.0045	0.0038

## Citation

If you find this paper useful in your research, please consider citing:

Contact

Thanks for your attention! If you have any suggestions or questions, you can leave a message here or contact us directly:

• yuliyunlut@hotmail.com
• xiaochen.yang.16@ucl.ac.uk
• mazhanyu@bupt.edu.cn
• xiaoxulilut@gmail.com