CoolFace
Datasetpublic

orianrivlin/fashion-lookmatch-dataset

👔 Fashion LookMatch Synthetic Dataset Overview This dataset contains 1,000 synthetic fashion images generated using Stable Diffusion XL (SDXL). It was created as part of the "Build Your Own AI Application" final project. The dataset is designed for building a Visual Retrieval & Outfit Recommendation System. 📊 Exploratory Data Analysis (EDA) To ensure the quality and balance of the dataset, we performed a rigorous EDA. 1. Category… See the full description on the dataset page: https://huggingface.co/datasets/orianrivlin/fashion-lookmatch-dataset.

sourceHugging Facemitupdated 8mo agoView on Hugging Face
0likes80downloads
Dataset Card

👔 Fashion LookMatch Synthetic Dataset

Overview

This dataset contains 1,000 synthetic fashion images generated using Stable Diffusion XL (SDXL). It was created as part of the "Build Your Own AI Application" final project. The dataset is designed for building a Visual Retrieval & Outfit Recommendation System.

📊 Exploratory Data Analysis (EDA)

To ensure the quality and balance of the dataset, we performed a rigorous EDA.

1. Category Balance

We ensured a uniform distribution across all 12 fashion categories to prevent class bias during retrieval. Category Distribution

2. Style Diversity

The dataset covers multiple style tags (Casual, Formal, Streetwear, etc.). Note that items can be multi-labeled. Style Distribution

3. Color Distribution

We analyzed the color spectrum to verify visual diversity. Color Distribution

4. Correlation Heatmap

We analyzed the relationship between Categories and Styles to understand the dataset's semantic structure. Heatmap

5. Data Samples

Below is a random sample of images demonstrating the "Product Photo" aesthetic and label accuracy. Sample Grid

📁 Dataset Structure

Each sample contains the following fields:

FieldTypeDescription
imageImageThe generated fashion item (1024x1024)
categoryStringE.g., 'jacket', 'sneakers', 'dress'
style_tagsList[Str]E.g., ['casual', 'vintage']
colorStringDominant color (optional)
seasonStringSeasonality (optional)
captionStringNatural language description
seedIntThe seed used for generation (Reproducibility)

🛠️ How to Use

python
from datasets import load_dataset

# Load the full dataset
dataset = load_dataset("orianrivlin/fashion-lookmatch-dataset", split="train")

# View the first example
print(dataset[0])

🤖 Generation Process

Model: stabilityai/stable-diffusion-xl-base-1.0

Constraints: "product photo, plain background, no logos"

Metadata: Structured prompt generation based on predefined schemas.