Research map / Label-efficient and robust learning / Domain adaptation and generalization
Bidirectional One-Shot Unsupervised Domain Mapping
Tomer Cohen, Lior Wolf
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 feature cycle loss f-cycle (ablation) | Ours without f-cycle (f-bab, f-aba on) · MNIST one-shot, B->A | accuracy: 26.20 -> 30.73 | Improved |
| add feature cycle loss f-cycle (ablation) | Ours without f-cycle · SVHN one-shot, B->A (MNIST to SVHN) | accuracy: 48.25 -> 69.48 | Improved |
| add f-bab (ablation) | Ours without f-bab (f-aba, f-cycle on) · MNIST one-shot, B->A | accuracy: 30.02 -> 30.73 | Improved |
| add f-aba (ablation) | Ours without f-aba (f-bab, f-cycle on) · MNIST one-shot, B->A | accuracy: 30.02 -> 30.73 | No clear effect |
| add f-aba (ablation) | Ours without f-aba · SVHN one-shot, B->A (MNIST to SVHN) | accuracy: 68.12 -> 69.48 | No clear effect |
| add f-bab (ablation) | Ours without f-bab · SVHN one-shot, B->A (MNIST to SVHN) | accuracy: 17.80 -> 69.48 | Improved |
| add all three cycle losses (ablation) | no cycle losses · MNIST one-shot, A->B (MNIST to SVHN) | accuracy: 39.00 -> 66.50 | Improved |
| add f-cycle only (ablation) | no cycle losses · MNIST one-shot, A->B (MNIST to SVHN) | accuracy: 39.00 -> 61.50 | Improved |
| add f-aba only (ablation) | no cycle losses · MNIST one-shot, A->B (MNIST to SVHN) | accuracy: 39.00 -> 36.50 | Hurt |
| replace encodings of B with random permutations of latent codes in f-cycle (ablation) | Ours · MNIST one-shot (A->B and B->A) | accuracy: 65.00,23.53,23.00,63.45 | Mixed |
Closest papers by meaning
- One-Shot Unsupervised Cross Domain Translation (NeurIPS 2018 · similarity 0.88)
- Unsupervised Cross-Domain Image Generation (ICLR 2017 · similarity 0.87)
- One-Sided Unsupervised Domain Mapping (NeurIPS 2017 · similarity 0.87)
- Unsupervised Domain Generalization by Learning a Bridge Across Domains (CVPR 2022 · similarity 0.86)
- Cycle-consistent Masked AutoEncoder for Unsupervised Domain Generalization (ICLR 2023 · similarity 0.86)
- Revisiting Continuity of Image Tokens for Cross-domain Few-shot Learning (ICML 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.
