Retail-71 is proposed in the paper "Robust Unsupervised Domain Adaptation Through Negative-View Regularization" accepted to WACV 2024.
This dataset can only be used for scientific/non-commercial purposes. If you use this dataset in your research, please cite the corresponding paper (Link).
You can use the following bibtex code.
@InProceedings{Jang_2024_WACV,
author = {Jang, Joonhyeok and Lee, Sunhyeok and Kim, Seonghak and Kim, Jung-un and Kim, Seonghyun and Kim, Daeshik},
title = {Robust Unsupervised Domain Adaptation Through Negative-View Regularization},
booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
month = {January},
year = {2024},
pages = {2462-2471}
}Note: The first author of this paper, Joonhyeok Jang, changed the name to Sehui Jang. (February 21th, 2025)
We introduce Retail-71 whose source domain contains clean images of diverse products commonly seen in market or convenience store, while its target domain broods images in which hand occlusion and motion blur exist. The domain shift between source and target domain is mainly from hand occlusion and motion blur, as well as a difference in background, resolution, lighting, device used to collect images, and so on.
Unlike other existing benchmarks, Retail-71 includes its test set with three difficulties, i.e. easy, medium, and hard level. Harder test set contains the test samples with more intensified motion blur and hand occlusion.
You can download Retail-71: Link (about 18.27GB).
When it comes to source domain, this dataset includes not only plain version (denoted as 0) but also augmented versions (denoted as 1, 2, 3, 4, 01, 02, ..., 0134, 01234) by rule-based synthesis (RS) . Note that target domain is NOT augmented by RS.
The meaning of number code in denotations of augmented versions is as follows.
- 0 means O: original source-domain dataset (i.e., no RS is applied)
- 1 means E: intermediate-domain dataset generated by RS with easy difficulty
- 2 means M: intermediate-domain dataset generated by RS with medium difficulty
- 3 means MP: intermediate-domain dataset generated by RS with medium difficulty and zero padding
- 4 means H: intermediate-domain dataset generated by RS with hard difficulty
- The multi-digit indicates the multiple intermediate datasets corresponding to the digits are mixed. Some examples are as follows.
- 013 means the mixture of O, E, and MP.
- 02 means the mixture of O and M.
The structure of Retail-71 is as follows.
Retail-71
|_ retail71 # train set
| |_ 0 # O: Its source-domain dataset is an original source-domain dataset (i.e., no RS is applied)
| | |_ train # source domain
| | | |_ 001 # class folder
| | | | |_ <name_image1>.jpg
| | | | |_ <name_image2>.jpg
| | | | ...
| | | | |_ <name_imageN>.jpg
| | | |_ 002
| | | ...
| | | |_ 071
| | | |_ train_image_list.txt
| | |_ validation # target domain
| | | |_ 001 # class folder
| | | |_ 002
| | | ...
| | | |_ 071
| | | |_ valid_image_list.txt
| | |_ category.txt
| |_ 1 # E: Its source-domain dataset is an intermediate-domain dataset generated by RS with EASY difficulty
| |_ 2 # M: Its source-domain dataset is an intermediate-domain dataset generated by RS with MEDIUM difficulty
| |_ 3 # MP: Its source-domain dataset is an intermediate-domain dataset generated by RS with MEDIUM difficulty and zero padding
| |_ 4 # H: Its source-domain dataset is an intermediate-domain dataset generated by RS with HARD difficulty
| ...
| |_ 0134 # Its source-domain dataset is the mixture of O, E, MP, and H
| |_ 01234 # Its source-domain dataset is the mixture of O, E, M, MP, and H
|_ test # test set
| |_ 1 # difficulty: easy
| | |_ train # It is an empty folder
| | |_ validation # Test samples are here.
| | |_ category.txt
| |_ 2 # difficulty: medium
| |_ 3 # difficulty: hard