Research map / Vision-language and multimodal / Human and embodied modeling
Effectively Leveraging Attributes for Visual Similarity
Samarth Mishra, Zhongping Zhang, Yuan Shen, Ranjitha Kumar, Venkatesh Saligrama, Bryan A. Plummer
Read the paperCheck 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 attribute supervision to similarity conditions | PAN-GE · Polyvore Outfits (resampled) | FITB accuracy: 64.1 -> 69.7 | Improved |
| add attribute supervision to similarity conditions | PAN · CUB-200-2011 (5-way 5-shot) | accuracy: 92.60 -> 92.77 | No clear effect |
| add attribute supervision to similarity conditions | ProxyNCA++ & PAN · In-Shop Clothes Retrieval | Recall@1: 91.4 -> 92.1 | Improved |
| predict attributes per image then classify similarity (Attr. Similarity) instead of CGAE | CGAE · Polyvore Outfits (original) | FITB accuracy: 74.1 -> 69.5 | Hurt |
| add attribute prediction by pairwise attribute similarity | ProxyNCA++ · In-Shop Clothes Retrieval | Recall@1: 90.9 -> 86.4 | Hurt |
| add multitask attribute head to Siamese network (single batch training) (ablation) | Siamese Network · CUB-200-2011 (5-way 5-shot) | accuracy: 89.01 -> 75.82 | Hurt |
| train on the whole training split in a single batch (ablation) | Siamese Network · Polyvore Outfits (resampled) | FITB accuracy: 36.4 -> 44.7 | Improved |
| train on the whole training split in a single batch (ablation) | Siamese Network · CUB-200-2011 (5-way 5-shot) | accuracy: 76.87 -> 89.01 | Improved |
| add graph encoder (GE) instead of CNN (ablation) | PAN-Supervised · Polyvore Outfits (resampled) | FITB accuracy: 28.2 -> 69.7 | Improved |
| add graph encoder (GE) instead of CNN (ablation) | PAN-Supervised · CUB-200-2011 (5-way 5-shot) | accuracy: 92.77 -> 90.16 | Hurt |
| supervise with XNOR instead of OR (ablation) | PAN-Supervised-GE · Polyvore Outfits (resampled) | FITB accuracy: 69.7 -> 60.8 | Hurt |
| supervise with XNOR instead of OR (ablation) | PAN-Supervised + ProxyNCA++ · In-Shop Clothes Retrieval | Recall@1: 91.5 -> 92.1 | Improved |
Closest papers by meaning
- Efficient Discovery and Effective Evaluation of Visual Perceptual Similarity: A Benchmark and Beyond (ICCV 2023 · similarity 0.89)
- Learning Similarity Conditions Without Explicit Supervision (ICCV 2019 · similarity 0.88)
- Towards Latent Attribute Discovery From Triplet Similarities (ICCV 2019 · similarity 0.88)
- Learning Attribute-Driven Disentangled Representations for Interactive Fashion Retrieval (ICCV 2021 · similarity 0.87)
- Context-Aware Visual Compatibility Prediction (CVPR 2019 · similarity 0.87)
- Interpretable Measures of Conceptual Similarity by Complexity-Constrained Descriptive Auto-Encoding (CVPR 2024 · 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.
