DeepFashion (Large Scale Fashion)
DeepFashion is a reference dataset in the field of fashion assisted by artificial intelligence. It brings together hundreds of thousands of annotated images of clothing, making it possible to develop applications for visual recognition, personalized recommendations or trend analysis.
Over 800,000 images in JPEG format with JSON annotations
Usable for academic research purposes only. License to be consulted for commercial uses
Description
The DeepFashion dataset includes:
- More than 800,000 JPEG images of clothing in a varied context (on a mannequin, isolated, in store...)
- JSON annotations for:
- Categories (dress, pants, jacket...)
- Style attributes (color, patterns, sleeves, etc.)
- Keypoints for alignment or adjustment
- Relationships between clothes (associated top/bottom, visual sets...)
DeepFashion covers a wide range of styles, cuts, and contexts of appearance of clothing, allowing an effective generalization of AI models on uses related to fashion.
What is this dataset for?
DeepFashion is mainly used for:
- Training models to detect, classify or recognize clothing
- The improvement of visual search or personalized recommendation systems in e-commerce
- Analysis of clothing trends and style segmentation
- Automatic detection and alignment of clothing in images or videos
Can it be enriched or improved?
Yes, despite its wealth, DeepFashion can be optimized by:
- The addition of more realistic contexts such as street scenes or moving shots
- The integration of new categories specific to certain cultures or emerging trends
- Coupling with text descriptions to create multimodal image/text models
- Collaborative or semi-supervised annotation of new points of visual interest (folds, fabrics, accessories)
🔗 Source: DeepFashion Dataset
Frequently Asked Questions
What types of models are most commonly trained with DeepFashion?
Models for multi-label classification, object detection, key point identification, and similarity networks for visual recommendation are the most common. GaNs are also used there for the generation of stylized images.
Does DeepFashion cover men's and women's fashion?
Yes, even though a majority of the images are oriented towards women's fashion, the dataset also includes mixed or masculine items. For an optimal balance, it can be combined with other specialized datasets.
How do I use DeepFashion to find products based on an image?
A visual similarity model can be trained using attribute annotations and element relationships. This type of template can then suggest similar items based on a user photo or reference image.
