Research map / Detection and segmentation / CNN features
Dual Cross-Attention Learning for Fine-Grained Visual Categorization and Object Re-Identification
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 GLCA and PWCA to DeiT-Tiny | DeiT-Tiny · CUB-200-2011 | accuracy: 82.1 -> 84.6 | Helped |
| Add GLCA and PWCA to ViT-Base | ViT-Base · MSMT17 | mAP: 61.6 -> 64.0 | Helped |
| Add GLCA only (ablation) | DeiT-Tiny · CUB-200-2011 | accuracy: 82.1 -> 83.1 | Helped |
| Add PWCA only (ablation) | DeiT-Tiny · CUB-200-2011 | accuracy: 82.1 -> 83.1 | Helped |
| Use global key-value vectors with selected local queries rather than local self-attention (ablation) | GLCA on DeiT-Tiny · CUB-200-2011 | accuracy: 82.6 -> 83.1 | Helped |
| Select local queries from aggregated attention responses rather than randomly (ablation) | GLCA on DeiT-Tiny · CUB-200-2011 | accuracy: 82.6 -> 83.1 | Helped |
| Select queries using aggregated attention rather than the penultimate layer alone (ablation) | GLCA on DeiT-Tiny · CUB-200-2011 | accuracy: 82.8 -> 83.1 | Helped |
| Use cross-token strategy instead of GLCA (ablation) | DeiT-Tiny · CUB-200-2011 | accuracy: 82.1 vs. 83.1 | Hurt |
| Add image noise to the target image (ablation) | DeiT-Tiny · CUB-200-2011 | accuracy: 82.1 -> 77.3 | Hurt |
| Add noise to the target label (ablation) | DeiT-Tiny · CUB-200-2011 | accuracy: 82.1 -> 81.6 | Hurt |
| Replace distractor image embeddings with Gaussian noise (ablation) | PWCA on DeiT-Tiny · CUB-200-2011 | accuracy: 83.1 -> 82.1 | Hurt |
| Sample the distractor from COCO rather than the same dataset (ablation) | PWCA on DeiT-Tiny · CUB-200-2011 | accuracy: 83.1 -> 82.5 | Hurt |
Closest papers by meaning
- Dual Attention Networks for Few-Shot Fine-Grained Recognition (AAAI 2022 · similarity 0.88)
- Looking for the Devil in the Details: Learning Trilinear Attention Sampling Network for Fine-Grained Image Recognition (CVPR 2019 · similarity 0.88)
- Look Closer to See Better: Recurrent Attention Convolutional Neural Network for Fine-Grained Image Recognition (CVPR 2017 · similarity 0.87)
- Exploring Self-Attention for Image Recognition (CVPR 2020 · similarity 0.87)
- ResMatch: Residual Attention Learning for Feature Matching (AAAI 2024 · similarity 0.86)
- Linear Differential Vision Transformer: Learning Visual Contrasts via Pairwise Differentials (NeurIPS 2025 · similarity 0.85)
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.