dress-code

Why we included this project

Image-based virtual try-on is a hard problem to benchmark, and this repository gives you both the data and the method to work on it. The Dress Code dataset holds more than 50,000 high-resolution garment and model image pairs split across dresses, upper-body, and lower-body clothing, with keypoints, skeletons, label maps, and dense poses included as auxiliary annotations. The repo also documents the multi-category try-on approach from the ECCV 2022 paper, so it works as a reference implementation for reproducing the method or comparing results against it. One thing to plan around: the dataset is gated behind a signed release agreement and only open to institutional researchers, so it is not a grab-and-go download. For teams building fashion e-commerce try-on features, this is a solid reference for both data and methodology.