Research map / Image generation / GANs
SketchyGAN: Towards Diverse and Realistic Sketch to Image Synthesis
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 SketchyGAN instead of pix2pix, Augmented+Label | Sketch-to-image synthesis · Sketchy test split, 50 categories | Inception Score: 5.49 -> 7.90 | Helped |
| Use SketchyGAN instead of pix2pix, Augmented+Label | Sketch-to-image synthesis · Sketchy test split, 50 categories | Picked as more realistic: 21.8% -> 53.7% | Helped |
| Use SketchyGAN instead of pix2pix, Augmented+Label | Sketch-to-image synthesis · Sketchy test split, 50 categories | Input correctly identified: 65.9% -> 47.4% | Hurt |
| Add edge-map image pairs to Sketchy training data | pix2pix · Sketchy test split, 50 categories | Inception Score: 3.94 -> 4.53 | Helped |
| Add class conditioning and auxiliary classification supervision | pix2pix trained on Augmented Sketchy · Sketchy test split, 50 categories | Inception Score: 4.53 -> 5.49 | Helped |
| Replace discrete edge pre-training and sketch fine-tuning with gradual edge-to-sketch training | SketchyGAN · Sketchy test split, 50 categories | Inception Score: 6.73 vs 7.90 | Helped |
| Add GAN loss and discriminator (ablation) | SketchyGAN · Sketchy test split, 50 categories | Inception Score: 1.49 -> 7.90 | Helped |
| Add generator class supervision and discriminator auxiliary classification loss (ablation) | SketchyGAN · Sketchy test split, 50 categories | Inception Score: 6.64 -> 7.90 | Helped |
| Add L1 and perceptual losses (ablation) | SketchyGAN · Sketchy test split, 50 categories | Inception Score: 6.70 -> 7.90 | Helped |
| Add diversity loss (ablation) | SketchyGAN · Sketchy test split, 50 categories | Inception Score: 7.29 -> 7.90 | Helped |
| Use MRU blocks instead of Improved ResNet blocks | Sketch-to-image generator and discriminator · Sketchy test split, 50 categories | Inception Score: 5.76 -> 7.90 | Helped |
| Use MRU blocks instead of CRN | Sketch-to-image generator · Sketchy test split, 50 categories | Inception Score: 4.56 -> 7.90 | Helped |
Closest papers by meaning
- SketchyCOCO: Image Generation From Freehand Scene Sketches (CVPR 2020 · similarity 0.89)
- SketchGAN: Joint Sketch Completion and Recognition With Generative Adversarial Network (CVPR 2019 · similarity 0.89)
- APDrawingGAN: Generating Artistic Portrait Drawings From Face Photos With Hierarchical GANs (CVPR 2019 · similarity 0.88)
- Creative Sketch Generation (ICLR 2021 · similarity 0.88)
- Spatial Fusion GAN for Image Synthesis (CVPR 2019 · similarity 0.87)
- Scribbler: Controlling Deep Image Synthesis With Sketch and Color (CVPR 2017 · 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.