Research map / Point clouds / Point cloud classification
Neural Point-Based Graphics
Kara-Ali Aliev, Artem Sevastopolsky, Maria Kolos, Dmitry Ulyanov, Victor Lempitsky
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 learnable neural descriptors (vs point color input, Pix2Pix) (ablation) | point-based renderer · ScanNet Studio | LPIPS: 0.535 -> 0.472 | Improved |
| add learnable neural descriptors (vs point color input, Pix2Pix) (ablation) | point-based renderer · People Person 2 | LPIPS: 0.1642 -> 0.1131 | Improved |
| add learnable neural descriptors (vs point color input, Pix2Pix) (ablation) | point-based renderer · ScanNet LivingRoom | FID: 192.82 -> 138.87 | Improved |
| use point cloud instead of mesh proxy (Ours vs DNR) | neural rendering · ScanNet Studio | FID: 84.92 -> 76.73 | Improved |
| use point cloud instead of mesh proxy (Ours vs DNR) | neural rendering · Plant | FID: 49.61 -> 41.67 | Improved |
| use point cloud instead of mesh proxy (Ours vs DNR) | neural rendering · Owl | FID: 54.14 -> 78.55 | Hurt |
| use point cloud instead of mesh proxy (Ours vs DNR) | neural rendering · ScanNet LivingRoom | FID: 129.33 -> 138.87 | Hurt |
| use point cloud instead of mesh proxy (Ours vs DNR) | neural rendering · Plant (user study) | user preference: 50.3% -> 69% | Improved |
| use point cloud instead of mesh proxy (Ours vs DNR) | neural rendering · ScanNet Studio (user study) | user preference: 49.7% vs. 50.3% | No clear effect |
| increase point cloud from small to large (ablation) | Ours, descriptor size 8 · ScanNet Studio | VGG loss: 632.39 -> 609.11 | Improved |
| increase descriptor size from 8 to 16 (ablation) | Ours, large cloud · ScanNet Studio | VGG loss: 609.11 -> 611.38 | No clear effect |
| increase descriptor size from 4 to 8 (ablation) | Ours, large cloud · ScanNet Studio | VGG loss: 610.76 -> 609.11 | Improved |
Closest papers by meaning
- Deep Novel View Synthesis from Colored 3D Point Clouds (ECCV 2020 · similarity 0.89)
- Ponder: Point Cloud Pre-training via Neural Rendering (ICCV 2023 · similarity 0.89)
- NPBG++: Accelerating Neural Point-Based Graphics (CVPR 2022 · similarity 0.88)
- Neural Point Cloud Diffusion for Disentangled 3D Shape and Appearance Generation (CVPR 2024 · similarity 0.86)
- HashPoint: Accelerated Point Searching and Sampling for Neural Rendering (CVPR 2024 · similarity 0.86)
- Point2Real: Bridging the Gap between Point Cloud and Realistic Image for Open-World 3D Recognition (AAAI 2024 · similarity 0.86)
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.
