Multi-Reference Virtual Try-On Fashion Images

DressCode-MR is a large-scale dataset containing over 28,000 multi-reference virtual try-on samples. It is constructed upon the original DressCode dataset to facilitate models handling multiple fashion items simultaneously. The dataset was created by zhengchong and last updated in early 2026.

Use Cases

• Train virtual try-on models to generate images of a person wearing multiple fashion items like tops, bottoms, and dresses from reference images.
• Evaluate model performance on multi-reference tasks involving simultaneous handling of tops, bottoms, shoes, and bags.
• Benchmark computer vision systems for fashion item compatibility and outfit generation using the 28,000+ multi-reference samples.

Strengths

• Contains over 28,000 multi-reference virtual try-on samples, providing substantial scale.
• Specifically designed for multi-item tasks, covering tops, bottoms, dresses, shoes, and bags.
• Built upon the established DressCode dataset, suggesting a structured foundation.

Limitations

• Specific sample details, image resolutions, and demographic or geographic coverage are not provided in the input.
• The dataset's construction method and potential biases from the original source are not described.

Provenance

Source

Hugging Face, created by author zhengchong.

Collection Method

Constructed upon the original DressCode dataset.

Time Range

null

Freshness

Last updated 2026-01-04.

Geography

null