Codatta/Fashion-1K

---
license: openrail
tags:
- fashion
- clothing
- virtual-try-on
- e-commerce
- flatlay
- image-generation
pretty_name: Fashion 1K
size_categories:
- 1K<n<10K
task_categories:
- object-detection
language:
- en
---

# Fashion 1K

## Dataset Summary

**Fashion 1K** is a curated collection of 1,000 high-quality fashion images, focusing on apparel and outfit compositions without human models.

Unlike typical street-style datasets (like DeepFashion) that include human poses and complex backgrounds, this dataset provides **clean, human-free** images. The images primarily feature **Flat Lay** (clothing arranged on a flat surface) or **Ghost Mannequin** styles, making them ideal for tasks that require a clear view of the garment's structure, texture, and color without occlusion.

**Key Features:**
* **Human-Free:** No faces, limbs, or skin tones—strictly focused on the garments.
* **Outfit-Centric:** Many images showcase complete looks (e.g., Top + Bottom + Shoes) to aid in compatibility learning.
* **Clean Backgrounds:** Minimized background noise to facilitate easier segmentation and feature extraction.

## Supported Tasks

This dataset is particularly suitable for:

* **Virtual Try-On (VTON):** Serving as the "garment" reference image (`g_img`) for 2D try-on pipelines.
* **Fashion Compatibility Learning:** Learning which items (e.g., shirt and trousers) go well together based on the curated outfits.
* **Generative AI Training:** Training LoRAs or ControlNets for specific clothing styles without the bias of human figures.
* **E-commerce Tagging:** Automated classification of clothing categories and attributes.

## Dataset Structure

### Data Fields

* **`image`** (image): The high-resolution image of the clothing item or outfit.

## Usage Example

```python
from datasets import load_dataset

# Load the dataset
ds = load_dataset("Codatta/Fashion-1K", split="train")

# Display the first image
sample = ds[0]
sample['image'].show()

许可证：OpenRail
标签：
- 时尚
- 服装
- 虚拟试穿（Virtual Try-On）
- 电子商务
- 平铺拍摄（Flat Lay）
- 图像生成
展示名称：Fashion 1K
样本量范围：1000 < n < 10000
任务类别：目标检测
语言：英语

# Fashion 1K

## 数据集摘要

**Fashion 1K** 是一套精心整理的1000张高质量时尚图像合集，聚焦于无人体模特的服饰单品与整体穿搭搭配。

与包含人体姿态与复杂背景的典型街拍风格数据集（如DeepFashion）不同，本数据集提供**无人体干扰**的高清图像。图像主要采用**平铺拍摄（Flat Lay）**（将衣物平铺于平面摆放）或**幽灵模特（Ghost Mannequin）** 风格，非常适合需要清晰观察衣物结构、纹理与色彩且无遮挡的任务。

### 关键特性
* **无人体元素**：无面部、四肢或肤色，仅聚焦服饰本身。
* **穿搭为核心**：大量图像展示完整穿搭组合（例如上衣+下装+鞋履），助力搭配兼容性学习任务。
* **背景简洁**：背景噪音极低，便于开展图像分割与特征提取任务。

## 支持任务

本数据集尤其适用于：

* **虚拟试穿（Virtual Try-On, VTON）**：作为2D试穿流程中的“服饰参考图像（`g_img`）”。
* **时尚搭配兼容性学习**：基于整理好的穿搭组合，学习服饰单品（例如衬衫与长裤）的搭配逻辑。
* **生成式AI（Generative AI）训练**：针对特定服饰风格训练LoRA或ControlNet，规避人体带来的偏见。
* **电子商务标签标注**：实现服饰类别与属性的自动化分类。

## 数据集结构

### 数据字段

* **`image`**（图像）：服饰单品或整体穿搭的高分辨率图像。

## 使用示例

python
from datasets import load_dataset

# 加载数据集
ds = load_dataset("Codatta/Fashion-1K", split="train")

# 展示第一张图像
sample = ds[0]
sample['image'].show()

搜集汇总

背景与挑战

背景概述

Fashion-1K 是一个包含1,000张高质量时尚图像的数据集，专注于服装和搭配，不含人类模特，采用平铺或幽灵模特风格。该数据集适用于虚拟试穿、时尚兼容性学习和生成AI训练等任务，特点是图像背景干净、无遮挡，便于服装结构分析。