Research map / 3D vision / 3D generation
Neural Wireframe Renderer: Learning Wireframe to Image Translations
Yuan Xue, Zihan Zhou, Xiaolei Huang
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 joint representation learning (wireframe decoder branch) (ablation) | Ours · Wireframe dataset indoor test (422 images) | FID: 97.49 -> 70.73 | Improved |
| add joint representation learning (wireframe decoder branch) (ablation) | Ours · Wireframe dataset indoor test (422 images) | LPIPS: 2.85 -> 2.77 | Improved |
| add joint representation learning (wireframe decoder branch) (ablation) | Ours · Wireframe dataset indoor test (422 images) | SSIM: 0.092 -> 0.102 | Improved |
| add joint representation learning (wireframe decoder branch) (ablation) | Ours · Wireframe dataset indoor test (422 images) | sAP5: 26.7 -> 60.1 | Improved |
| add joint representation learning (wireframe decoder branch) (ablation) | Ours · Wireframe dataset indoor test (422 images) | sAP10: 34.4 -> 64.1 | Improved |
| add joint representation learning (wireframe decoder branch) (ablation) | Ours · Wireframe dataset indoor test (422 images) | sAP15: 37.5 -> 65.7 | Improved |
| use multi-scale discriminators and adversarial perceptual loss (pix2pixHD over pix2pix) | pix2pix · Wireframe dataset indoor test (422 images) | FID: 186.91 -> 153.36 | Improved |
| use pretrained-VGG perceptual loss without joint representation instead of pix2pixHD's learned perceptual loss | pix2pixHD · Wireframe dataset indoor test (422 images) | FID: 153.36 -> 97.49 | Improved |
| use full model instead of SPADE | SPADE · Wireframe dataset indoor test (422 images) | FID: 93.90 -> 70.73 | Improved |
| use full model instead of SPADE | SPADE · Wireframe dataset indoor test (422 images) | sAP5: 54.7 -> 60.1 | Improved |
| use full model instead of pix2pixHD in human preference | pix2pixHD · Human preference study (422 test images) | preference rate (%): 3.7% -> 65.1% | Improved |
| add joint wireframe/image generation to StackGANv2 | StackGANv2 · Noise-to-image generation (wireframe dataset) | Inception score: 2.92 -> 3.08 | Improved |
Closest papers by meaning
- Learning to Parse Wireframes in Images of Man-Made Environments (CVPR 2018 · similarity 0.90)
- NEAT: Distilling 3D Wireframes from Neural Attraction Fields (CVPR 2024 · similarity 0.89)
- Generating 3D House Wireframes with Semantics (ECCV 2024 · similarity 0.88)
- Learning to Reconstruct 3D Manhattan Wireframes From a Single Image (ICCV 2019 · similarity 0.87)
- EdgeDiff: Edge-aware Diffusion Network for Building Reconstruction from Point Clouds (CVPR 2025 · similarity 0.84)
- Wired Perspectives: Multi-View Wire Art Embraces Generative AI (CVPR 2024 · similarity 0.83)
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.
