Research map / Label-efficient and robust learning / Domain adaptation and generalization
Asking Friendly Strangers: Non-Semantic Attribute Transfer
Nils Murrugarra-Llerena, Adriana Kovashka
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 |
|---|---|---|---|
| replace learned attention weights with uniform weights (multi-domain, DD) (ablation) | ATTENTION-DD · overall average | F-measure: 0.67 -> 0.08 | Hurt |
| replace learned attention weights with uniform weights (all domains, AD) (ablation) | ATTENTION-AD · overall average | F-measure: 0.67 -> 0.17 | Hurt |
| replace learned attention weights with uniform weights (single domain) (ablation) | ATTENTION-SD · overall average | accuracy: 0.92 -> 0.80 | Hurt |
| add non-semantic transfer from other domains (DD) over target-only | TARGET-ONLY · overall average | F-measure: 0.62 -> 0.67 | Improved |
| add non-semantic transfer from other domains (DD) over target-only | TARGET-ONLY · textures | F-measure: 0.68 -> 0.96 | Improved |
| add non-semantic transfer from other domains (DD) over target-only | TARGET-ONLY · objects | F-measure: 0.50 -> 0.47 | Hurt |
| add non-semantic transfer from other domains (DD) over target-only | TARGET-ONLY · overall average | accuracy: 0.91 -> 0.94 | Improved |
| transfer from all domains including target domain (AD) instead of other domains only (DD) | ATTENTION-DD · overall average | F-measure: 0.67 -> 0.67 | No clear effect |
| attention transfer vs Tzeng et al. domain confusion (DD) | CONFUSION-DD · overall average | F-measure: 0.63 -> 0.67 | Improved |
| attention transfer vs Tzeng et al. domain confusion (DD) | CONFUSION-DD · overall average | accuracy: 0.94 -> 0.94 | No clear effect |
| attention transfer vs standard fine-tuning (AD) | FINETUNE-AD · overall average | F-measure: 0.39 -> 0.67 | Improved |
| ensemble ATTENTION-SD with TARGET-ONLY | ATTENTION-SD · overall average | F-measure: 0.62 -> 0.64 | Improved |
Closest papers by meaning
- Joint Semantic and Latent Attribute Modelling for Cross-Class Transfer Learning (TPAMI 2017 · similarity 0.88)
- Understanding Degeneracies and Ambiguities in Attribute Transfer (ECCV 2018 · similarity 0.88)
- Automatic Discovery, Association Estimation and Learning of Semantic Attributes for a Thousand Categories (CVPR 2017 · similarity 0.87)
- Improving Closed and Open-Vocabulary Attribute Prediction Using Transformers (ECCV 2022 · similarity 0.86)
- Attributes as Operators: Factorizing Unseen Attribute-Object Compositions (ECCV 2018 · similarity 0.85)
- Tree of Attributes Prompt Learning for Vision-Language Models (ICLR 2025 · similarity 0.84)
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.
