Research map / Image generation / Editing and style transfer
TSIT: A Simple and Versatile Framework for Image-to-Image Translation
Liming Jiang, Changxu Zhang, Mingyang Huang, Chunxiao Liu, Jianping Shi, Chen Change Loy
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 content stream (CS) (ablation) | TSIT · BDD100K multi-modal | FID: 89.429 -> 85.876 | Improved |
| add style stream (SS) (ablation) | TSIT · BDD100K multi-modal | FID: 86.263 -> 85.876 | Improved |
| add FADE feature transformation (ablation) | TSIT · BDD100K multi-modal | FID: 86.463 -> 85.876 | Improved |
| add FAdaIN feature transformation (ablation) | TSIT · BDD100K multi-modal | FID: 89.795 -> 85.876 | Improved |
| keep full channel width in CS/SS instead of dividing by 4 (ablation) | TSIT · day->night style transfer | FID: 95.199 -> 79.697 | Improved |
| keep full channel width instead of dividing by 2 (ablation) | TSIT · Cityscapes semantic synthesis | mIoU: 61.0 -> 65.9 | Improved |
| keep full channel width instead of dividing by 4 (ablation) | TSIT · Cityscapes semantic synthesis | FID: 74.4 -> 59.2 | Improved |
| keep full channel width instead of dividing by 4 (ablation) | TSIT · BDD100K multi-modal | FID: 97.297 -> 85.876 | Improved |
| feature-level injection (FAdaIN) instead of image-level AdaIN (ablation) | TSIT · day->night style transfer | FID: 80.618 -> 79.697 | Improved |
| feature-level injection (FADE) instead of image-level SPADE (ablation) | TSIT · Cityscapes semantic synthesis | mIoU: 59.7 -> 65.9 | Improved |
| feature-level injection (FADE) instead of image-level SPADE (ablation) | TSIT · Cityscapes semantic synthesis | FID: 60.1 -> 59.2 | Improved |
| TSIT versus DMIT on summer->winter | TSIT · summer->winter | FID: 87.969 -> 80.138 | Improved |
Later papers that compared with this method
4 later accepted papers report TSIT as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- A Diffusion Model Translator for Efficient Image-to-Image Translation (TPAMI 2024 · better on 5 of 6 shared results)
- SHUNIT: Style Harmonization for Unpaired Image-to-Image Translation (AAAI 2023 · better on 3 of 4 shared results)
- Generating Aligned Pseudo-Supervision From Non-Aligned Data for Image Restoration in Under-Display Camera (CVPR 2023 · better on 1 of 1 shared result)
- Enhancing Photorealism Enhancement (TPAMI 2022 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- TraVeLGAN: Image-To-Image Translation by Transformation Vector Learning (CVPR 2019 · similarity 0.89)
- StEP: Style-Based Encoder Pre-Training for Multi-Modal Image Synthesis (CVPR 2021 · similarity 0.88)
- Toward Multimodal Image-to-Image Translation (NeurIPS 2017 · similarity 0.88)
- Encoding in Style: A StyleGAN Encoder for Image-to-Image Translation (CVPR 2021 · similarity 0.88)
- LANIT: Language-Driven Image-to-Image Translation for Unlabeled Data (CVPR 2023 · similarity 0.88)
- MARS: Mixture of Auto-Regressive Models for Fine-grained Text-to-image Synthesis (AAAI 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.
