We present the FashionStyle14 dataset which focuses on predicting the fashion style of images. The images focus on single individuals with fully visible poses. We provide expert-curated fashion style annotations for a total of 14 unique challenging classes that focus on modern Japanese fashion styles such as Gal, Natural, or Casual.

For a related datasets, please see Fashion144k, Fashion144k (Stylenet), and Fashion550k.

• Download (418 MiB)

Overview

This is the dataset release that accompanies the paper:

What Makes a Style: Experimental Analysis of Fashion Prediction
Moeko Takagi, Edgar Simo-Serra, Satoshi Iizuka, Hiroshi Ishikawa
International Conference on Computer Vision Workshops (ICCVW), 2017

If you use this data, please cite it as:

@InProceedings{TakagiICCVW2017,
  author    = {Moeko Takagi and Edgar Simo-Serra and Satoshi Iizuka and Hiroshi Ishikawa},
  title     = {{What Makes a Style: Experimental Analysis of Fashion Prediction}},
  booktitle = "Proceedings of the International Conference on Computer Vision Workshops (ICCVW)",
  year      = 2017,
}

Layout

The images are split into different folders by class. Additionally, csv files for the train-validation-test splits used in the paper are provided.

README.md # This readme
dataset/ # All the images split in folders by class
train.csv # Train split files
val.csv # Validation split files
test.csv # Test split files

Note that we encourage to use this dataset for evaluating style classification and other applications as a generalization test set and not for just supervised training.