Research map / 3D vision / 3D generation
NEOMAP: Novel-View Synthesis via Noise Initialization by Manifold Alternating Projection
Jinxi Li, Tianyi Zhang, Yafei YANG, Zihui Zhang, Peng Huang, Koon Lin, Bo Yang
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 the full NeoMap alternating projection pipeline on top of reversed noise (ablation) | Wan2.2-I2V-A14B · Tanks-and-Temples ablation | FID-2048: 160.994 -> 83.341 | Improved |
| add pixel-constrained projection (PCP) (ablation) | NeoMap · Tanks-and-Temples ablation | ATE: 0.089 -> 0.008 | Improved |
| add PCP in pixel space instead of latent space (ablation) | NeoMap · Tanks-and-Temples ablation | FID-2048: 93.343 -> 83.341 | Improved |
| add anchored manifold projection (AMP) (ablation) | NeoMap · Tanks-and-Temples ablation | FID-2048: 91.215 -> 83.341 | Improved |
| increase alternating iterations from 5 to 15 (ablation) | NeoMap · Tanks-and-Temples ablation | FID-2048: 96.359 -> 83.341 | Improved |
| raise similarity threshold lambda from 0.7 to 0.8 (ablation) | NeoMap · Tanks-and-Temples ablation | FID-2048: 92.013 -> 83.341 | Improved |
| raise similarity threshold lambda from 0.8 to 0.9 (ablation) | NeoMap · Tanks-and-Temples ablation | ATE: 0.008 -> 0.022 | Hurt |
| add trajectory re-anchoring for 3 steps (ablation) | NeoMap · Tanks-and-Temples ablation | ATE: ATE 0.017 -> 0.008; FID-2048 83.750 -> 83.341 | Mixed |
| extend re-anchoring from 3 to 8 steps (ablation) | NeoMap · Tanks-and-Temples ablation | FID-2048: 83.341 -> 96.721 | Hurt |
| add PCP in the re-anchoring stage (ablation) | NeoMap · Tanks-and-Temples ablation | FID-2048: 83.341 -> 107.733 | Hurt |
| use NeoMap instead of stepwise-guidance NVS-Solver | Wan2.2-I2V-A14B · DAVIS (dynamic monocular video NVS) | FVD: 341.127 -> 234.634 | Improved |
| use Wan2.2 backbone instead of Wan2.1 (ablation) | NeoMap · Tanks-and-Temples (static single image NVS) | FID-2048: 89.527 -> 83.341 | Improved |
Closest papers by meaning
- NAMVIS: Next-Scale Autoregressive Multi-View Image Synthesis (NeurIPS 2026 · similarity 0.88)
- Dynamic View Synthesis as an Inverse Problem (NeurIPS 2025 · similarity 0.88)
- MultiDiff: Consistent Novel View Synthesis from a Single Image (CVPR 2024 · similarity 0.88)
- NVComposer: Boosting Generative Novel View Synthesis with Multiple Sparse and Unposed Images (CVPR 2025 · similarity 0.88)
- Novel View Synthesis from A Few Glimpses via Test-Time Natural Video Completion (NeurIPS 2025 · similarity 0.88)
- Reconstruct, Inpaint, Test-Time Finetune: Dynamic Novel-view Synthesis from Monocular Videos (NeurIPS 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.
