Research map / Label-efficient and robust learning / Domain adaptation and generalization
LANIT: Language-Driven Image-to-Image Translation for Unlabeled Data
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 Top-3 activated attributes instead of Top-1 (ablation) | LANIT · CelebA-HQ | mFID: 49.65 -> 41.68 | Helped |
| Add the slack domain to Top-3 LANIT (ablation) | LANIT · CelebA-HQ | mFID: 41.68 -> 33.39 | Helped |
| Add prompt learning to LANIT with Top-3 and slack domain (ablation) | LANIT · CelebA-HQ | mFID: 33.39 -> 32.71 | Helped |
| Use Top-3 activated attributes instead of Top-1 (ablation) | LANIT · CelebA-HQ | D&C: 0.561 / 0.320 -> 0.681 / 0.338 | Helped |
| Add the slack domain to Top-3 LANIT (ablation) | LANIT · CelebA-HQ | D&C: 0.681 / 0.338 -> 0.963 / 0.366 | Mixed |
| Add prompt learning to LANIT with Top-3 and slack domain (ablation) | LANIT · CelebA-HQ | D&C: 0.963 / 0.366 -> 1.181 / 0.425 | Helped |
| Use dataset-level text prompts and multi-hot labeling rather than unsupervised clustering | LANIT vs. TUNIT · Animal Faces-10 | mFID: 47.7 -> 32.67 | Helped |
| Use dataset-level text prompts and multi-hot labeling rather than unsupervised clustering | LANIT vs. TUNIT · Food-10 | mFID: 52.2 -> 49.13 | Helped |
| Increase the number of candidate domain descriptions (ablation) | LANIT · Animal Faces-10, Food-10, CelebA-HQ | mFID: Animal Faces-10: 68.7 (N=4), 47.9 (N=7), 32.7 (N=10), 32.9 (N=13), 34.3 (N=16); Food-10: 61.2 (N=4), 50.5 (N=7), 49.1 (N | Mixed |
| Add the slack domain (ablation) | LANIT · Animal Faces-10, Food-10, CelebA-HQ | Acc. | Helped |
| Learn and refine the domain prompts during training (ablation) | LANIT · Animal Faces-10, Food-10 | Acc. | Helped |
| Use multiple activated attributes for images with multiple attributes (ablation) | LANIT · CelebA-HQ | mFID: 49.65 -> 41.68 | Helped |
Closest papers by meaning
- Semi-Supervised Learning for Few-Shot Image-to-Image Translation (CVPR 2020 · similarity 0.87)
- Exemplar Guided Unsupervised Image-to-Image Translation with Semantic Consistency (ICLR 2019 · similarity 0.87)
- Towards Identifiable Unsupervised Domain Translation: A Diversified Distribution Matching Approach (ICLR 2024 · similarity 0.86)
- MuLan: Adapting Multilingual Diffusion Models for Hundreds of Languages with Negligible Cost (ICML 2025 · similarity 0.85)
- Leveraging Unpaired Data for Vision-Language Generative Models via Cycle Consistency (ICLR 2024 · similarity 0.84)
- Target Semantics Clustering via Text Representations for Robust Universal Domain Adaptation (AAAI 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.