Research map / Architectures and efficiency / Attention and transformers
TransFG: A Transformer Architecture for Fine-Grained Recognition
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 |
|---|---|---|---|
| Apply TransFG rather than ViT | ViT-B_16 · CUB-200-2011 | accuracy: 90.3 -> 91.7 | Helped |
| Apply TransFG rather than ViT | ViT-B_16 · Stanford Cars | accuracy: 93.7 -> 94.8 | Helped |
| Apply TransFG rather than ViT | ViT-B_16 · Stanford Dogs | accuracy: 91.7 -> 92.3 | Helped |
| Apply TransFG rather than ViT | ViT-B_16 · NABirds | accuracy: 89.9 -> 90.8 | Helped |
| Apply TransFG rather than ViT | ViT-B_16 · iNat2017 | accuracy: 68.7 -> 71.7 | Helped |
| Use overlapping rather than non-overlapping patches (ablation) | ViT · CUB-200-2011 | accuracy: 90.3 -> 90.5 | Helped |
| Use overlapping rather than non-overlapping patches (ablation) | TransFG · CUB-200-2011 | accuracy: 91.5 -> 91.7 | Helped |
| Use overlapping rather than non-overlapping patches (ablation) | TransFG · CUB-200-2011 | training time: 1.98 -> 5.38 | Hurt |
| Add Part Selection Module (ablation) | ViT · CUB-200-2011 | accuracy: 90.3 -> 91.0 | Helped |
| Add contrastive loss (ablation) | ViT · CUB-200-2011 | accuracy: 90.3% to 90.7% | Helped |
| Add contrastive loss (ablation) | TransFG · CUB-200-2011 | accuracy: 91.0% to 91.5% | Helped |
| Set contrastive-loss margin alpha to 0.2 rather than 0 (ablation) | TransFG · CUB-200-2011 | accuracy: 91.1 -> 91.4 | Helped |
Closest papers by meaning
- A Simple Interpretable Transformer for Fine-Grained Image Classification and Analysis (ICLR 2024 · similarity 0.89)
- Democratizing Fine-grained Visual Recognition with Large Language Models (ICLR 2024 · similarity 0.87)
- ViTree: Single-Path Neural Tree for Step-Wise Interpretable Fine-Grained Visual Categorization (AAAI 2024 · similarity 0.87)
- Vision Transformers Need Registers (ICLR 2024 · similarity 0.87)
- Linear Differential Vision Transformer: Learning Visual Contrasts via Pairwise Differentials (NeurIPS 2025 · similarity 0.87)
- Analyzing and Boosting the Power of Fine-Grained Visual Recognition for Multi-modal Large Language Models (ICLR 2025 · similarity 0.87)
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.