Research map / Image generation / Editing and style transfer
DRIT++: Diverse Image-to-Image Translation via Disentangled Representations
Hsin-Ying Lee, Hung-Yu Tseng, Qi Mao, Jia‐Bin Huang, Yu-Ding Lu, Maneesh Singh, Ming–Hsuan Yang
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 discriminator (ablation) | DRIT · Yosemite winter to summer | FID: 46.92 -> 41.34 | Improved |
| add latent regression loss (ablation) | DRIT · Yosemite winter to summer | FID: 53.12 -> 41.34 | Improved |
| add latent regression loss (ablation) | DRIT · Yosemite winter to summer | LPIPS: 0.0158 -> 0.0965 | Improved |
| add KL loss (ablation) | DRIT · Yosemite winter to summer | FID: 40.08 -> 41.34 | Hurt |
| add mode-seeking regularization (DRIT++) | DRIT · Yosemite winter to summer | LPIPS: 0.0965 -> 0.1183 | Improved |
| add mode-seeking regularization (DRIT++) | DRIT · Cat to Dog | LPIPS: 0.245 -> 0.280 | Improved |
| add mode-seeking regularization (DRIT++) | DRIT · Cat to Dog | FID: 24.306 -> 17.253 | Improved |
| add content discriminator (ablation) | DRIT++ · cat2dog | content distance: 55.45 -> 10.45 | Improved |
| add content discriminator (ablation) | DRIT++ · Yosemite | content distance: 58.69 -> 31.56 | Improved |
| use multi-domain model instead of two-domain | DRIT++ · Yosemite winter to summer | FID: 41.02 -> 44.86 | Hurt |
| use multi-domain model instead of two-domain | DRIT++ · Yosemite winter to summer | LPIPS: 0.1183 -> 0.1204 | Improved |
| add multi-scale generator-discriminator instead of two more layers (ablation) | DRIT++ high-resolution · GTA5 to Cityscape | FID: 37.19 -> 28.62 | Improved |
Later papers that compared with this method
20 later accepted papers report DRIT as a baseline in their results tables; 19 of them report a better number on most of the results they share with it.
- Physics-Informed Guided Disentanglement in Generative Networks (TPAMI 2023 · better on 3 of 3 shared results)
- BBDM: Image-to-Image Translation With Brownian Bridge Diffusion Models (CVPR 2023 · better on 0 of 2 shared results)
- Unpaired Image Translation via Vector Symbolic Architectures (ECCV 2022 · better on 4 of 4 shared results)
- Quality Metric Guided Portrait Line Drawing Generation From Unpaired Training Data (TPAMI 2022 · better on 4 of 4 shared results)
- Unpaired Cartoon Image Synthesis via Gated Cycle Mapping (CVPR 2022 · better on 2 of 2 shared results)
- MCL-GAN: Generative Adversarial Networks with Multiple Specialized Discriminators (NeurIPS 2022 · better on 2 of 2 shared results)
- Instance-Wise Hard Negative Example Generation for Contrastive Learning in Unpaired Image-to-Image Translation (ICCV 2021 · better on 3 of 3 shared results)
- TransferI2I: Transfer Learning for Image-to-Image Translation From Small Datasets (ICCV 2021 · better on 3 of 3 shared results)
12 more not listed.
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- StarGAN — StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translation (CVPR 2018) · better on 1 of 2 shared results
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
- Diverse Image-to-Image Translation via Disentangled Representations (ECCV 2018 · similarity 0.97)
- Multi-mapping Image-to-Image Translation via Learning Disentanglement (NeurIPS 2019 · similarity 0.90)
- Scaling-Up Disentanglement for Image Translation (ICCV 2021 · similarity 0.89)
- Image-to-image translation for cross-domain disentanglement (NeurIPS 2018 · similarity 0.88)
- SynthRGB-T: Language-Vision Guided Image Translation for Diversity Synthesis (CVPR 2026 · similarity 0.87)
- Transferring disentangled representations: bridging the gap between synthetic and real images (NeurIPS 2024 · 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.
