2,078 3D garment models across 10 categories including tops, skirts, and pants, paired with multi-view images and point clouds. The dataset provides ground-truth 3D geometry and feature lines for reconstructing clothing from single-view inputs.
Use Cases
- Train single-view 3D reconstruction models using the multi-view image pairs and ground-truth 3D meshes
- Develop garment segmentation and classification algorithms using the 10 category labels
- Perform point cloud completion or surface refinement using the dense point cloud data and feature line annotations
Strengths
- 2,078 3D garment models reconstructed from real-world scans
- 10 distinct garment categories including dresses, coats, and trousers
- Includes multi-view images and dense point clouds for each garment
- Provides semantic feature lines and 3D landmarks for geometric alignment