IndoFashion Dataset¶
Indofashion dataset is the first large-scale ethnic dataset of over 106K images with 15 different categories for fine-grained classification of Indian ethnic clothes. For more information, refer to the paper. To access the dataset, please fill out this form. We will provide you script to download the dataset.

Dataset Description¶
Dataset Statistics¶
Our dataset consists of 106K images and 15 unique cloth categories. For a fair evaluation, we ensure equal distribution of the classes in the validation and the test set consisting of 500 samples per class. The dataset stats are listed below.
Table 1: Class Categories.
| Gender | Categories |
| Women | Saree, Women Kurta, Leggings & Salwar, Palazzo, Lehenga, Dupatta, Blouse, Gown, Dhoti Pants, Petticoats, Women Mojari |
| Men | Men Kurta, Nehru Jacket, Sherwani, Men Mojari |
Table 2: Dataset stats.
| Split | # Images |
| Train | 91,166 |
| Valid | 7,500 |
| Test | 7,500 |
Data Format¶
In the Indofashion dataset, training, validation and test sets are provided as JSON (JavaScript Object Notation) text files with the following attributes for every data sample stored as a dictionary:
File Structure for train.json, val.json and test.json
{ "image_url": <image_url>,
"image_path": <img_path>,
"brand": <brand_name>,
"product_title": <product_title>,
"class_label": <class_label>
}
image_url: URL used to download the imageimage_path: Source path in dataset directory for the imagebrand: Brand name of the productproduct_title: Brief description of the productclass_label: Label of the product
*In cases where any of these attributes are not present, we substitute them with NA.
Citation¶
If you find our dataset or paper useful for research , please include the following citation:
@misc{rajput2021indofashion,
title={IndoFashion : Apparel Classification for Indian Ethnic Clothes},
author={Pranjal Singh Rajput and Shivangi Aneja},
year={2021},
eprint={2104.02830},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
Contact Us¶
For any queries, please email us at indofashion.dataset@gmail.com. We will try to respond as soon as possible.