The data used for experiments in the paper can be downloaded via following instructions.
- Download the data by running
download_data.sh
(it might take time to download the data due to its size (5GB)). - Once the data is ready, you may proceed to EXPERIMENTS.md.
We provide two csv files under the data directory, which are:
- train.csv : contains path and label information of each sample in the training set.
- filter_train.csv: contains path and label information of filtered images by representative sampling method described in supplementary material.
- test.cvs: contains path and label information of each sample in the test set
The given csv files will be used in experiments to train models.