Research map / Label-efficient and robust learning / Recognition and long-tail
Texture and Shape Biased Two-Stream Networks for Clothing Classification and Attribute Recognition
Yuwei Zhang, Peng Zhang, Chun Yuan, Zhi Wang
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 |
|---|---|---|---|
| Add landmark joint learning (ablation) | VGG16 clothing classifier and attribute recognizer · DeepFashion-C | category top-3 accuracy: 89.94 -> 91.54 | Improved |
| Add landmark joint learning (ablation) | VGG16 attribute recognizer · DeepFashion-C, shape attributes | top-3 recall: 60.03 -> 61.70 | Improved |
| Use ImageNet pre-training (ablation) | VGG16 attribute recognizer · DeepFashion-C, texture attributes | top-3 recall: 55.73 -> 58.34 | Improved |
| Use ImageNet pre-training (ablation) | VGG16 attribute recognizer · DeepFashion-C, shape attributes | top-3 recall: 60.03 -> 60.41 | No clear effect |
| Add landmark-aware attention to landmark joint learning (ablation) | VGG16 clothing classifier · DeepFashion-C | category top-3 accuracy: 91.59 -> 91.56 | No clear effect |
| Combine landmark joint learning and ImageNet pre-training in one stream (ablation) | VGG16 attribute recognizer · DeepFashion-C | all-attribute top-3 recall: 49.23 -> 48.77 | Hurt |
| Separate texture-biased and shape-biased features into two streams (ablation) | VGG16 attribute recognizer · DeepFashion-C | all-attribute top-3 recall: 48.77 -> 50.58 | Improved |
| Jointly train category classification and attribute recognition (ablation) | VGG16 multitask model · DeepFashion-C | category top-3 accuracy: 87.64 -> 89.94 | Improved |
| Jointly train category classification and attribute recognition (ablation) | VGG16 multitask model · DeepFashion-C, shape attributes | top-3 recall: 58.50 -> 60.03 | Improved |
| Add landmark joint learning (ablation) | clothing retrieval attribute recognizer · In-shop Clothes Retrieval | top-30 recall: 75.29 -> 75.33 | No clear effect |
| Add landmark joint learning (ablation) | clothing retrieval attribute recognizer · Consumer-to-Shop Clothes Retrieval | top-20 recall: 65.29 -> 68.42 | Improved |
| Use the proposed two-stream TS-FashionNet | clothing classifier and attribute recognizer · DeepFashion-C test | all-attribute top-3 recall: 49.19 -> 50.58 | Improved |
Closest papers by meaning
- Fashion Landmark Detection in the Wild (ECCV 2016 · similarity 0.88)
- Attentive Fashion Grammar Network for Fashion Landmark Detection and Clothing Category Classification (CVPR 2018 · similarity 0.87)
- Layout-Graph Reasoning for Fashion Landmark Detection (CVPR 2019 · similarity 0.87)
- DeepFashion: Powering Robust Clothes Recognition and Retrieval With Rich Annotations (CVPR 2016 · similarity 0.87)
- Learning Attribute and Class-Specific Representation Duet for Fine-Grained Fashion Analysis (CVPR 2023 · similarity 0.86)
- AnyDressing: Customizable Multi-Garment Virtual Dressing via Latent Diffusion Models (CVPR 2025 · similarity 0.84)
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.
