24679-Project-1 / README.md
danjung9's picture
Update README.md
8fa9af2 verified
metadata
dataset_info:
  features:
    - name: image
      dtype: image
    - name: approval(1-10)
      dtype: int64
    - name: category
      dtype: string
  splits:
    - name: original
      num_bytes: 6612290
      num_examples: 524
    - name: augmented
      num_bytes: 884638915.5
      num_examples: 11004
  download_size: 891157080
  dataset_size: 891251205.5
configs:
  - config_name: default
    data_files:
      - split: original
        path: data/original-*
      - split: augmented
        path: data/augmented-*

Dataset Summary

This dataset contains over 11,500 annotated images of fashion apparel and accessories, designed for machine learning research in fashion understanding, style recommendation, and approval rating prediction. Each image is labeled with:

  • A category label (from 62 distinct clothing or accessory types)
  • An approval rating (integer, 1–10 scale)

Images are provided in a consistent format and resolution, making the dataset suitable for computer vision applications such as image classification, regression, and multi-task learning.

Supported Tasks and Benchmarks

  • Image Classification (category prediction)
  • Regression (approval rating prediction)
  • Multi-label/Multi-task Learning (joint category and rating prediction)
  • Fashion Recommendation and Outfit Rating

Example use cases:

  • Training models for fashion item classification
  • Building automated approval/rating predictors
  • Benchmarking deep learning architectures on fashion-related tasks

Dataset Structure

Field Type Description
image image Apparel or accessory item image
approval (1–10) int64 Human-assigned rating (1–10)
category string Category name (62 possible values)
  • File format: Parquet
  • Resolution: 228x228 px

Data Splits

  • original: 524 rows (likely original photos)
  • augmented: 11,000+ rows (expanded via augmentation techniques)

Class Balance and Diversity

  • Categories: 62 unique fashion types (T-Shirt, Jeans, Skirt, Handbag, Jacket, etc.)
  • Approval Ratings: Well-distributed across 1–10
  • Image Variability: Organic and augmented images for robustness

Dataset Size

  • Total Rows: 11,528
  • Total Size: ~891 MB

Languages

  • Category labels and documentation in English

Usage

Please cite or link this dataset in your work when using it for academic, research, or commercial purposes.

from datasets import load_dataset
ds = load_dataset("danjung9/24679-Project-1")

Citation

If you use this dataset, please cite as:

@dataset{danjung9_24679_project_1,
  title = {Fashion Approval and Category Dataset},
  author = {danjung9},
  year = {2025},
  url = {https://huggingface.co/datasets/danjung9/24679-Project-1}
}