Research map / Reinforcement learning / Visual and representation RL
Paint Transformer: Feed Forward Neural Painting With Stroke Prediction
Songhua Liu, Tianwei Lin, Dongliang He, Fu Li, Ruifeng Deng, Xin Li, Errui Ding, Hao Wang
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 pixel loss (ablation) | Paint Transformer · Real images (100 landscapes, 100 WikiArt, 100 FFHQ) | pixel loss: 0.081 -> 0.056 | Improved |
| add pixel loss (ablation) | Paint Transformer · Real images (100 landscapes, 100 WikiArt, 100 FFHQ) | perceptual loss: 1.012 -> 0.807 | Improved |
| add parameter L1 loss (ablation) | Paint Transformer · Synthesized random strokes | stroke parameter L1 distance D_L1: 0.267 -> 0.083 | Improved |
| add parameter L1 loss (ablation) | Paint Transformer · Real images (100 landscapes, 100 WikiArt, 100 FFHQ) | pixel loss: 0.074 -> 0.056 | Improved |
| add Wasserstein loss (ablation) | Paint Transformer · Synthesized random strokes | Wasserstein distance D_W: 0.034 -> 0.018 | Improved |
| add Wasserstein loss (ablation) | Paint Transformer · Real images (100 landscapes, 100 WikiArt, 100 FFHQ) | pixel loss: 0.069 -> 0.056 | Improved |
| add confidence (BCE) loss (ablation) | Paint Transformer · Real images (100 landscapes, 100 WikiArt, 100 FFHQ) | pixel loss: 0.071 -> 0.056 | Improved |
| add confidence (BCE) loss (ablation) | Paint Transformer · Synthesized random strokes | pixel loss: 0.052 -> 0.042 | Improved |
| use Paint Transformer instead of Optim | neural painting · Real images (100 landscapes, 100 WikiArt, 100 FFHQ) | pixel loss: 0.059 -> 0.056 | Improved |
| use Paint Transformer instead of RL | neural painting · Real images (100 landscapes, 100 WikiArt, 100 FFHQ) | pixel loss: 0.040 -> 0.056 | Hurt |
| use Paint Transformer instead of Optim for stroke prediction | neural painting · Synthesized random strokes | stroke parameter L1 distance D_L1: 0.137 -> 0.083 | Improved |
| use Paint Transformer instead of Optim | neural painting · Inference time, 512x512 image | inference time (s): 521.447 -> 0.304 | Improved |
Closest papers by meaning
- Learning to Paint With Model-Based Deep Reinforcement Learning (ICCV 2019 · similarity 0.91)
- StrokeNet: A Neural Painting Environment (ICLR 2019 · similarity 0.90)
- Neural 3D Strokes: Creating Stylized 3D Scenes with Vectorized 3D Strokes (CVPR 2024 · similarity 0.86)
- MindPainter: Efficient Brain-Conditioned Painting of Natural Images via Cross-Modal Self-Supervised Learning (AAAI 2025 · similarity 0.85)
- Goal Conditioned Reinforcement Learning for Photo Finishing Tuning (NeurIPS 2024 · similarity 0.84)
- PrefPaint: Aligning Image Inpainting Diffusion Model with Human Preference (NeurIPS 2024 · 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.
