Research map / Image generation / Editing and style transfer
Dynamic Instance Normalization for Arbitrary Style Transfer
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 a MobileNet-based network with DIN instead of existing arbitrary stylization architectures | Arbitrary style transfer · 512 × 512 input | Computational cost: 14.16 -> 3.62 encoder GFLOPs; 19.59 -> 3.72 decoder GFLOPs | Helped |
| Replace standard convolutions with depthwise separable convolutions (ablation) | DIN network · Style transfer ablation | Visual quality | No clear effect |
| Use multiple hierarchical DIN layers instead of one single DIN layer (ablation) | DIN network · Style transfer ablation | Content structure preservation | Helped |
| Use deformable convolution in place of standard convolution (ablation) | DIN · Arbitrary style transfer | Spatial stroke control | Helped |
| Set the dynamic convolution kernel size to the input feature map size (ablation) | DIN for non-natural images · Non-natural images | Stroke generation in uniform pixel areas | Helped |
| Use DIN instead of AdaIN under the same lightweight network settings | Lightweight arbitrary style transfer · Training | Convergence | Helped |
| Use a separate trainable style encoder rather than shared content and style VGG encoders (ablation) | AdaIN · Arbitrary style transfer | Visual quality | Hurt |
| Replace shared VGG encoders with smaller shared encoders (ablation) | AdaIN · Arbitrary style transfer | Visual quality | Hurt |
| Use DIN with the same VGG encoder as other ASPM methods | Arbitrary style transfer · Challenging style patterns and fine details | Visual quality | Helped |
| Reduce the encoder to the MobileNet-based architecture while using DIN | Arbitrary style transfer · Style transfer quality | Visual quality | No clear effect |
Closest papers by meaning
- Adaptive Convolutions for Structure-Aware Style Transfer (CVPR 2021 · similarity 0.91)
- Adjustable Real-time Style Transfer (ICLR 2020 · similarity 0.89)
- A Content Transformation Block for Image Style Transfer (CVPR 2019 · similarity 0.88)
- StyleStudio: Text-Driven Style Transfer with Selective Control of Style Elements (CVPR 2025 · similarity 0.87)
- HSI: A Holistic Style Injector for Arbitrary Style Transfer (CVPR 2025 · similarity 0.87)
- SigStyle: Signature Style Transfer via Personalized Text-to-Image Models (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.