RichWear Dataset - Fashion AI

• 該資料由 資訊分析與經濟效果研究實驗室提供
• http://www.im.ntu.edu.tw/~lu/

RichWear dataset contains 321,698 street fashion images collected from wear.jp spanning from 2017 to 2019. All images are of size 500x667 pixels. Each image is associated with a date, gender of the character in the image, user-provided clothing brands, hashtags, colors, and categories. We also provide manually verified clothing patterns, colors, and categories of 3,868 images.

This dataset is suitable for analyzing street fashion trends or training fashion attribute recognition models.

If you are using this dataset, please cite our work:

Fu-Hsien Huang, Hsin-Min Lu, and Yao-Wen Hsu, (2021, Mar.). From Street Photos to Fashion Trends: Leveraging User-Provided Noisy Labels for Fashion Understanding, IEEE Access, Forthcoming, https://doi.org/10.1109/ACCESS.2021.3069245

使用平台上之資源所衍生之研究成果與論文發表，建議致謝內容範例：

We thank the NCHC Sci-DM project for data service and technical support. The project is funded by the Ministry of Science and Technology (MOST) ( MOST 108-2634-F-492-001).