Research map / Image generation / Editing and style transfer
Unsupervised Sketch to Photo Synthesis
Runtao Liu, Qian Yu, Stella X. Yu
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 |
|---|---|---|---|
| use two-stage (shape then content) design instead of one-stage cycleGAN (ablation) | cycleGAN · ShoeV2 | FID: 79.35 -> 51.80 | Improved |
| use two-stage design instead of one-stage cycleGAN (ablation) | cycleGAN · ChairV2 | FID: 177.79 -> 109.46 | Improved |
| use grayscale instead of edge map as intermediate target (ablation) | Ours · ShoeV2 | FID: 96.58 -> 48.73 | Improved |
| use grayscale instead of edge map as intermediate target (ablation) | Ours · ChairV2 | FID: 236.38 -> 100.51 | Improved |
| add self-supervised objective with noise sketch composition (ablation) | Ours · ShoeV2 (shape stage) | FID: 47.0 -> 46.46 | Improved |
| add self-supervised objective (ablation) | Ours · ChairV2 (shape stage) | FID: 93.33 -> 90.87 | Improved |
| add attention module (ablation) | Ours · ChairV2 (shape stage) | FID: 92.03 -> 90.87 | Improved |
| add attention module (ablation) | Ours · ShoeV2 (shape stage) | FID: 47.0 -> 46.46 | Improved |
| add both components over vanilla model (ablation) | Ours · ChairV2 (shape stage) | FID: 104.0 -> 90.87 | Improved |
| train unpaired instead of using paired Pix2Pix (ablation) | Ours · ShoeV2 (shape stage) | FID: 75.84 -> 46.46 | Improved |
| train unpaired instead of using paired Pix2Pix (ablation) | Ours · ChairV2 (shape stage) | FID: 164.01 -> 90.87 | Improved |
| remove latent pairing by merging ShoeV2 sketches with UT Zappos photos (ablation) | Ours · ShoeV2 | FID: 48.7 -> 48.6 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- UGATIT — U-GAT-IT: Unsupervised Generative Attentional Networks with Adaptive Layer-Instance Normalization for Image-to-Image Translation (ICLR 2020) · better on 4 of 4 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
- Picture That Sketch: Photorealistic Image Generation From Abstract Sketches (CVPR 2023 · similarity 0.94)
- Learning to Sketch With Shortcut Cycle Consistency (CVPR 2018 · similarity 0.92)
- Sketch Down the FLOPs: Towards Efficient Networks for Human Sketch (CVPR 2025 · similarity 0.90)
- It's All About Your Sketch: Democratising Sketch Control in Diffusion Models (CVPR 2024 · similarity 0.88)
- One-Shot Reference-based Structure-Aware Image to Sketch Synthesis (AAAI 2025 · similarity 0.88)
- SketchFusion: Learning Universal Sketch Features through Fusing Foundation Models (CVPR 2025 · similarity 0.88)
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.
