Research map / Image generation / Fast diffusion sampling
RenderFlow: Single-Step Neural Rendering via Flow Matching
Shenghao Zhang, Runtao Liu, Christopher Schroers, Yang Zhang
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 sparse keyframe guidance adapter (ablation) | RenderFlow · Rendering test set (512x512) | PSNR: 24.214 -> 26.663 | Improved |
| single-step flow model vs multi-step diffusion baselines | neural rendering · Rendering test set (512x512) | time per frame (s): 1.40 -> 0.19 | Improved |
| RenderFlow (w/o keyframe) vs deferred rendering | RenderFlow · Rendering test set (512x512) | PSNR: 24.649 -> 24.214 | Hurt |
| infer in a single step instead of 4 steps for a 4-timestep SDE-trained model (ablation) | RenderFlow · Rendering ablation (256x256) | PSNR: 23.384 -> 23.590 | Improved |
| use SDE bridge noise instead of ODE (1-step inference) (ablation) | RenderFlow · Rendering ablation (256x256) | PSNR: 23.304 -> 23.590 | Improved |
| use 4 fixed timesteps instead of uniform timesteps in SDE training (ablation) | RenderFlow · Rendering ablation (256x256) | PSNR: 22.192 -> 23.384 | Improved |
| add LPIPS loss to latent loss (ablation) | RenderFlow · Rendering ablation (256x256) | PSNR: 21.588 -> 23.538 | Improved |
| add gradient loss (ablation) | RenderFlow · Rendering ablation (256x256) | PSNR: 23.538 -> 23.590 | No clear effect |
| dedicated query instead of reused query in keyframe adapter (ablation) | Keyframe adapter · Rendering ablation (256x256) | PSNR: 24.520 -> 25.217 | Improved |
| add LoRA in feed-forward layer of keyframe adapter (ablation) | Keyframe adapter · Rendering ablation (256x256) | PSNR: 25.217 -> 25.720 | Improved |
| combine keyframe adapter with VACE progressive inference (ablation) | Keyframe adapter · Rendering ablation (256x256) | PSNR: 25.720 -> 26.220 | Improved |
| progressive inference only (no keyframe adapter) (ablation) | RenderFlow · Rendering ablation (256x256) | PSNR: 23.590 -> 25.042 | Improved |
Closest papers by meaning
- Diffusion Renderer: Neural Inverse and Forward Rendering with Video Diffusion Models (CVPR 2025 · similarity 0.91)
- Uni-Renderer: Unifying Rendering and Inverse Rendering Via Dual Stream Diffusion (CVPR 2025 · similarity 0.88)
- Contextual Flow Matching: Adaptive Step Selection in Flow Models for Efficient Visual Generation (NeurIPS 2026 · similarity 0.88)
- UMAMI: Unifying Masked Autoregressive Models and Deterministic Rendering for View Synthesis (NeurIPS 2025 · similarity 0.88)
- Improving Progressive Generation with Decomposable Flow Matching (NeurIPS 2025 · similarity 0.88)
- CAD: Photorealistic 3D Generation via Adversarial Distillation (CVPR 2024 · 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.
