tolgayan/fashion-styles

Fashion Styles

A structured taxonomy of fashion style labels for outfit analysis, visual style classification, retrieval, and LLM-based style judging.

The dataset contains 294 canonical style records. Each record pairs a human-readable style name and description with practical recognition metadata: visual indicators, color logic, silhouettes, common contexts, cultural or regional anchors, aesthetic moods, formality, mainstreamness, temporal references, and classifier guidance.

This is not an image dataset. It is a text and metadata knowledge base intended to support systems that need consistent fashion style labels and evidence-aware descriptions.

Dataset Configurations

The default config is styles.

Main Fields

The styles config starts with:

• —style_name: canonical style label, such as 90s Minimalism or Abaya Fashion
• —style_description: concise natural-language definition of the style
• —key_indicators: visual cues that help identify the style
• —color_schema: common palettes, color contrasts, or color families
• —silhouette: typical shape, fit, and proportion cues
• —style_family: broad style families, such as streetwear, modestwear, or retro_vintage
• —style_origin_logic: why the style exists as a category, such as subculture, historical period, occasion, material, or region
• —wear_context: common real-world settings where the style appears
• —dominant_visual_driver: the evidence a model or human judge should rely on most
• —classifier_role: guidance on whether the label is suitable as a primary label, secondary influence, umbrella category, fallback, or edge case

Most taxonomy fields are multi-label arrays because a style can naturally span several families, contexts, moods, or visual drivers.

Example Usage

from datasets import load_dataset

styles = load_dataset("tolgayan/fashion-styles", "styles")
metadata = load_dataset("tolgayan/fashion-styles", "metadata_fields")

first_style = styles["train"][0]
print(first_style["style_name"])
print(first_style["style_description"])
print(first_style["key_indicators"])

Load the schema dictionary:

field_docs = load_dataset("tolgayan/fashion-styles", "metadata_fields")

for field in field_docs["train"]:
    print(field["field_name"], field["description"])

Intended Uses

This dataset is useful for:

• —building fashion style classifiers or retrieval indexes
• —designing label spaces for outfit-tagging systems
• —grounding LLM prompts for style judging or outfit analysis
• —normalizing fashion style names across products, editorial systems, or annotation workflows
• —mapping style labels to broader metadata such as formality, context, mood, and visual evidence

Limitations

Style classification is inherently subjective and context-dependent. The records describe visible clothing and styling cues, not a wearer's identity, background, or intent. Cultural and regional labels should be applied only when there is clear visual evidence in the clothing or styling.

The taxonomy is designed for style reasoning and classification support. It should not be treated as a complete history of fashion, a demographic ontology, or a source for inferring personal attributes.

Citation

If you use this dataset, please cite this Hugging Face dataset and acknowledge the fashion style taxonomy it provides.