Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Bi-directional Feature Reconstruction Network for Fine-Grained Few-Shot Image Classification
Read the paperCheck 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 |
|---|---|---|---|
| Use bi-directional feature reconstruction with FSRM and FMRM | Bi-FRN with Conv-4 · CUB, Dogs, and Cars 5-way 1-shot/5-shot | accuracy: 79.08±0.20 vs 76.55±0.21 (CUB 1-shot); 92.22±0.10 vs 90.33±0.11 (CUB 5-shot) | Helped |
| Use bi-directional feature reconstruction with FSRM and FMRM | Bi-FRN with ResNet-12 · CUB and Cars 5-way 1-shot/5-shot | accuracy: 85.44±0.18 vs 85.15±0.18 (CUB 1-shot); 97.49±0.05 vs 97.02±0.06 (Cars 5-shot) | Helped |
| Use bi-directional feature reconstruction with FSRM and FMRM | Bi-FRN with ResNet-12 · Dogs 5-way 1-shot and 5-shot | accuracy: 76.89±0.21 vs 78.02±0.20; 88.27±0.12 vs 89.85±0.11 | Hurt |
| Add FSRM to FMRM (ablation) | Bi-FRN with Conv-4 · CUB, Dogs, and Cars 5-way 1-shot/5-shot | accuracy: CUB 1-shot: 74.92±0.21 -> 79.08±0.20; Dogs 1-shot: 61.28±0.21 -> 64.74±0.22; Dogs 5-shot: 80.07±0.14 -> 81.29±0.14; Cars | Mixed |
| Add FSRM to FMRM (ablation) | Bi-FRN with ResNet-12 · CUB, Dogs, and Cars 5-way 1-shot/5-shot | accuracy: CUB 1-shot: 84.33±0.18 -> 85.44±0.18; CUB 5-shot: 94.25±0.09 -> 94.73±0.09; Dogs 1-shot: 76.29±0.21 -> 76.89±0.21; Dogs | Mixed |
| Add FMRM to FSRM (ablation) | Bi-FRN with Conv-4 · CUB, Dogs, and Cars 5-way 1-shot/5-shot | accuracy: CUB 1-shot: 75.37±0.21 -> 79.08±0.20; CUB 5-shot: 88.61±0.12 -> 92.22±0.10; Dogs 1-shot: 65.10±0.22 -> 64.74±0.22; Dogs | Helped |
| Add FMRM to FSRM (ablation) | Bi-FRN with ResNet-12 · CUB, Dogs, and Cars 5-way 1-shot/5-shot | accuracy: CUB 1-shot: 82.53±0.19 -> 85.44±0.18; CUB 5-shot: 92.43±0.10 -> 94.73±0.09; Dogs 1-shot: 75.64±0.21 -> 76.89±0.21; Dogs | Mixed |
| Add support-to-query reconstruction to query-to-support reconstruction (ablation) | Bi-FRN with Conv-4 · CUB, Dogs, and Cars 5-way 1-shot/5-shot | accuracy: CUB 1-shot: 79.88±0.20 -> 79.08±0.20; CUB 5-shot: 91.76±0.11 -> 92.22±0.10; Dogs 1-shot: 65.26±0.22 -> 64.74±0.22; Dogs | Mixed |
| Add query-to-support reconstruction to support-to-query reconstruction (ablation) | Bi-FRN with Conv-4 · CUB, Dogs, and Cars 5-way 1-shot/5-shot | accuracy: CUB 1-shot: 76.54±0.21 -> 79.08±0.20; CUB 5-shot: 88.03±0.14 -> 92.22±0.10; Dogs 1-shot: 64.39±0.22 -> 64.74±0.22; Dogs | Helped |
| Add support-to-query reconstruction to query-to-support reconstruction (ablation) | Bi-FRN with ResNet-12 · CUB, Dogs, and Cars 5-way 1-shot/5-shot | accuracy: CUB 1-shot: 83.72±0.19 -> 85.44±0.18; CUB 5-shot: 93.31±0.09 -> 94.73±0.09; Dogs 1-shot: 76.50±0.21 -> 76.89±0.21; Dogs | Helped |
| Add query-to-support reconstruction to support-to-query reconstruction (ablation) | Bi-FRN with ResNet-12 · CUB, Dogs, and Cars 5-way 1-shot/5-shot | accuracy: CUB 1-shot: 81.72±0.19 -> 85.44±0.18; CUB 5-shot: 90.83±0.11 -> 94.73±0.09; Dogs 1-shot: 75.62±0.22 -> 76.89±0.21; Dogs | Helped |
Closest papers by meaning
- Few-Shot Fine-Grained Image Classification with Progressively Feature Refinement and Continuous Relationship Modeling (AAAI 2025 · similarity 0.88)
- Cross-Layer and Cross-Sample Feature Optimization Network for Few-Shot Fine-Grained Image Classification (AAAI 2024 · similarity 0.88)
- Twofold Debiasing Enhances Fine-Grained Learning with Coarse Labels (AAAI 2025 · similarity 0.86)
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (AAAI 2024 · similarity 0.86)
- Simple Semantic-Aided Few-Shot Learning (CVPR 2024 · similarity 0.86)
- Strong Baselines for Parameter-Efficient Few-Shot Fine-Tuning (AAAI 2024 · 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.