language:
- en
pretty_name: Virtual Try-On – Garment and Model Photo Dataset
license: cc-by-nc-2.0
Leverage our Virtual Try-On dataset to train your models for realistic garment transfer.
This is a sample dataset. To access the full version or request any custom dataset tailored to your needs, contact DataHive at [email protected].
The Virtual Try-On Dataset is a scalable, high-quality dataset designed to support the development and evaluation of AI models in the fashion and retail technology space. Each record represents a unique pairing of a human model and garment, offering high-resolution imagery from multiple perspectives along with detailed metadata. This dataset supports realistic visual diversity across clothing types, styles, and demographic categories, making it ideal for training virtual try-on systems, garment recognition models, and generative image tasks.
While this free trial set includes 5,000 models, the full dataset currently spans close to 300,000 model-garment pairs, with regular updates. Each sample may include multiple photographic views of both the model and garment, offering diverse visual contexts.
The dataset is delivered as a plain CSV file containing metadata and direct image URLs. Images are hosted on a scalable cloud infrastructure and can be accessed via script or browser.
Dataset Description
- Access: Free sample dataset
- Curated by: https://datahive.ai
- Language(s) (NLP): English
- License: Creative Commons Non-Commercial 4.0 (CC BY-NC 4.0)
Uses
- Train deep learning pipelines to simulate how garments look on different body shapes, genders, and poses.
- Develop models to generate new images of garments on varied model types and poses. Reduces dependency on expensive photo shoots and expands training coverage.
Dataset Structure
The dataset is delivered as a plain CSV file containing metadata and direct image URLs. Images are hosted on a scalable cloud infrastructure and can be accessed via script or browser.
Source Data
The dataset was curated using data and images gathered from publicly accessible websites. All content was obtained from open web sources that did not require authentication, paywalls, or proprietary access. Efforts were made to ensure that the data is either in the public domain or falls under fair use for research and trial purposes