Research map / Label-efficient and robust learning / Domain adaptation and generalization
Fine-Grained Recognition in the Wild: A Multi-Task Domain Adaptation Approach
Timnit Gebru, Judy Hoffman, Li Fei-Fei
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 classifiers and attribute consistency loss (ablation) | Source CNN · Cars->GSV unsupervised adaptation (all GSV val classes) | accuracy (%): 9.28 -> 14.37 | Improved |
| add attribute classifiers without consistency loss (ablation) | Source CNN · Cars->GSV unsupervised adaptation (all GSV val classes) | accuracy (%): 9.28 -> 10.80 | Improved |
| add consistency loss on top of attribute classifiers (ablation) | Source CNN w/att · Cars->GSV unsupervised adaptation (all GSV val classes) | accuracy (%): 10.80 -> 14.37 | Improved |
| add attribute losses and consistency loss to domain confusion (ablation) | DC · Cars->GSV unsupervised adaptation (all GSV val classes) | accuracy (%): 14.98 -> 19.05 | Improved |
| add attribute losses and consistency loss to semi-supervised DC (ablation) | DC · Cars->GSV semi-supervised adaptation (held-out GSV classes) | accuracy (%): 12.34 -> 19.11 | Improved |
| add attributes and consistency loss to S+T CNN (ablation) | S+T CNN · Cars->GSV semi-supervised adaptation (held-out GSV classes) | accuracy (%): 4.12 -> 7.45 | Improved |
| add WordNet-label classifiers without consistency loss (ablation) | Source CNN · Office Amazon->Webcam unsupervised | accuracy (%): 60.9 -> 59.5 | Hurt |
| add WordNet-label classifiers and consistency loss (ablation) | Source CNN · Office Amazon->Webcam unsupervised | accuracy (%): 60.9 -> 61.2 | No clear effect |
| add WordNet-label losses and consistency loss to DC (ablation) | DC · Office Amazon->Webcam unsupervised | accuracy (%): 61.1 -> 62.4 | Improved |
| add WordNet-label losses and consistency loss to semi-supervised DC (ablation) | DC · Office Amazon->Webcam semi-supervised | accuracy (%): 47.0 -> 51.8 | Improved |
| add WordNet-label classifiers and consistency loss to S+T CNN (ablation) | S+T CNN · Office Amazon->Webcam semi-supervised | accuracy (%): 45.5 -> 45.3 | No clear effect |
| add domain confusion without attributes | Source CNN · Cars->GSV unsupervised adaptation (all GSV val classes) | accuracy (%): 9.28 -> 14.98 | Improved |
Closest papers by meaning
- Webly-Supervised Fine-Grained Visual Categorization via Deep Domain Adaptation (TPAMI 2016 · similarity 0.91)
- The Unreasonable Effectiveness of Noisy Data for Fine-Grained Recognition (ECCV 2016 · similarity 0.91)
- Webly Supervised Fine-Grained Recognition: Benchmark Datasets and an Approach (ICCV 2021 · similarity 0.90)
- Fine-Grained Image Retrieval via Dual-Vision Adaptation (AAAI 2026 · similarity 0.89)
- Democratizing Fine-grained Visual Recognition with Large Language Models (ICLR 2024 · similarity 0.88)
- 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.
