Research map / Detection and segmentation / Open-set and few-shot detection
DeepFashion2: A Versatile Benchmark for Detection, Pose Estimation, Segmentation and Re-Identification of Clothing Images
Yuying Ge, Ruimao Zhang, Xiaogang Wang, Xiaoou Tang, Ping Luo
Check your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Combine pose and class features for retrieval rather than using class features alone (ablation) | Match R-CNN · DeepFashion2 validation retrieval, ground-truth boxes | top-1 accuracy: 0.122 -> 0.268 | Improved |
| Combine pose and class features for retrieval rather than using class features alone (ablation) | Match R-CNN · DeepFashion2 validation retrieval, detected boxes | top-20 accuracy: 0.417 -> 0.617 | Improved |
| Combine pose and class features rather than mask and class features (ablation) | Match R-CNN · DeepFashion2 validation retrieval, ground-truth boxes | top-1 accuracy: 0.214 -> 0.268 | Improved |
| Combine pose and class features rather than mask and class features (ablation) | Match R-CNN · DeepFashion2 validation retrieval, detected boxes | top-20 accuracy: 0.564 -> 0.617 | Improved |
| Use ground-truth boxes instead of detected boxes | Match R-CNN retrieval · DeepFashion2 validation | top-20 accuracy: 0.617 -> 0.665 | Improved |
| Evaluate landmark AP on visible and occluded landmarks rather than visible landmarks only | Mask R-CNN · DeepFashion2 validation, overall | AP_pt: 0.641 -> 0.563 | Hurt |
| Use pose features rather than mask features for retrieval (ablation) | Match R-CNN · DeepFashion2 validation retrieval, ground-truth boxes | top-1 accuracy: 0.187 -> 0.255 | Improved |
| Use pose features rather than mask features for retrieval (ablation) | Match R-CNN · DeepFashion2 validation retrieval, detected boxes | top-20 accuracy: 0.529 -> 0.589 | Improved |
| Evaluate detection on heavily occluded rather than slightly occluded items | Mask R-CNN · DeepFashion2 validation | AP_box: 0.712 -> 0.372 | Hurt |
| Evaluate detection on large zoom-in rather than no zoom-in items | Mask R-CNN · DeepFashion2 validation | AP_box: 0.695 -> 0.466 | Hurt |
| Evaluate segmentation on heavily occluded rather than slightly occluded items | Mask R-CNN · DeepFashion2 validation | AP_mask: 0.720 -> 0.389 | Hurt |
| Evaluate landmark estimation on heavily occluded rather than slightly occluded items | Mask R-CNN · DeepFashion2 validation, visible landmarks | AP_pt: 0.669 -> 0.398 | Hurt |
Closest papers by meaning
- DeepFashion: Powering Robust Clothes Recognition and Retrieval With Rich Annotations (CVPR 2016 · similarity 0.95)
- Deep Fashion3D: A Dataset and Benchmark for 3D Garment Reconstruction from Single Images (ECCV 2020 · similarity 0.90)
- FashionTailor: Controllable Clothing Editing for Human Images with Appearance Preserving (AAAI 2025 · similarity 0.86)
- MV-VTON: Multi-View Virtual Try-On with Diffusion Models (AAAI 2025 · similarity 0.86)
- ChatGarment: Garment Estimation, Generation and Editing via Large Language Models (CVPR 2025 · similarity 0.86)
- Robust-MVTON: Learning Cross-Pose Feature Alignment and Fusion for Robust Multi-View Virtual Try-On (CVPR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
