Outfit-Level Virtual Try-On Dataset with Multi-Reference Garments

80,000 outfit pairs link multiple reference garment images to a model wearing the complete look. ArtmeScienceLab created this dataset for high-fidelity virtual try-on research, with a test set released in March 2026. Each pair includes 3 to 12 reference images, averaging 4.48 items per outfit.

Use Cases

• Train image-to-image models to generate a model wearing an outfit from 3-12 reference garment images.
• Benchmark multi-reference virtual try-on systems using the provided outfit pairs and model images.
• Develop outfit compatibility algorithms using the 40 major and 300+ fine-grained clothing categories.
• Create conditional image generation models that synthesize a person wearing specified accessories and garments.

Strengths

• 80,000 many-garments-to-one-look pairs provide substantial training data.
• Each outfit pair includes an average of 4.48 reference garment images.
• Covers 40 major clothing categories and over 300 fine-grained subcategories.

Limitations

• Dataset size, specific row counts for the test set, and column details are unknown.
• Potential geographic or demographic bias in model images is unspecified.
• The description is a partial summary, requiring a visit to the external page for full details.

Provenance

Source

ArtmeScienceLab via Hugging Face.

Collection Method

null

Time Range

null

Freshness

Last updated March 2026.

Geography

null