CFPD 2013: Comprehensive Fashion Parsing DatasetTranslation site
An essential resource for research in fashion image analysis. Introduced in the paper 'Fashion Parsing with Weak Color-Category Labels' by Liu et al., and published in...
Tags:DatasetDataset Overview
The CFPD 2013 dataset comprises two main parts:
- Chictopia.com Images
- Images: 97,490 images
- Tags: 292,541 tags
- Source: Chictopia.com
- Description: These images come from Chictopia.com, a fashion community website. They include image-level tags that identify different fashion attributes and categories, offering a rich resource for understanding fashion trends and styles.
- Annotated Images
- Images: 2,682 images
- Pixel Annotations: Each image is annotated at the pixel level with both color and category labels.
- Color Labels: 13 distinct colors
- Category Labels: 23 different categories
- Description: This part of the dataset provides detailed annotations, making it suitable for tasks that require precise identification and classification of fashion elements.
Weakly Supervised Setting
The dataset supports a weakly supervised learning environment, where only image-level tags are available during the training phase. This makes it ideal for developing algorithms that can learn from limited supervision, a common scenario in real-world applications.
Applications in E-Commerce
The CFPD 2013 dataset is particularly valuable for e-commerce applications, such as:
- Product Recognition: Improving the accuracy of product identification in images, helping customers find similar items.
- Attribute Prediction: Enhancing search functionalities by predicting attributes like color, style, and category from product images.
- Inventory Management: Assisting in organizing and categorizing large inventories of fashion items more efficiently.
- Customer Personalization: Enabling better personalization by understanding customer preferences based on visual attributes of the products they interact with.
Citation
For referencing the CFPD 2013 dataset, please use the following citation:
@article{liu2013fashion,
title={Fashion parsing with weak color-category labels},
author={Liu, Si and Feng, Jiashi and Domokos, Csaba and Xu, Hui and Huang, Junshi and Hu, Zhenzhen and Yan, Shuicheng},
journal={IEEE Transactions on Multimedia},
volume={16},
number={1},
pages={253--265},
year={2013},
publisher={IEEE}
}
By leveraging the CFPD 2013 dataset, e-commerce platforms can significantly enhance their capabilities in fashion product recognition and customer experience, driving better engagement and sales.