Research map / 3D vision / 3D generation
Geometry-Aware Deep Network for Single-Image Novel View Synthesis
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 plane-based homographies and learned selection maps rather than only warping depth predictions | Geometry-aware network · KITTI | pixel ℓ1 error: 0.668 -> 0.340 | Helped |
| Use plane-based homographies and learned selection maps rather than appearance flow | Geometry-aware network · KITTI | pixel ℓ1 error: 0.471 -> 0.340 | Helped |
| Use plane-based homographies and learned selection maps rather than only warping depth predictions | Geometry-aware network · ScanNet | pixel ℓ1 error: 0.217 -> 0.167 | Helped |
| Add the refinement network (ablation) | Ours-Geo · KITTI | pixel ℓ1 error: 0.340 -> 0.345 | Hurt |
| Add the refinement network (ablation) | Ours-Geo · ScanNet | pixel ℓ1 error: 0.167 -> 0.176 | Hurt |
| Add learned selection maps rather than using seed regions as hard masks (ablation) | Geometry-aware network with ground-truth depth/normals · KITTI | pixel ℓ1 error: 0.357 -> 0.329 | Helped |
| Add learned selection maps rather than using seed regions as hard masks (ablation) | Geometry-aware network with estimated depth/normals · KITTI | pixel ℓ1 error: 0.373 -> 0.340 | Helped |
| Use ground-truth depth/normals instead of estimated depth/normals (ablation) | Geometry-aware network with learned selection maps · KITTI | pixel ℓ1 error: 0.340 -> 0.329 | Helped |
| Add learned selection maps rather than using seed regions as hard masks (ablation) | Geometry-aware network with ground-truth depth · ScanNet | pixel ℓ1 error: 0.174 -> 0.159 | Helped |
| Add learned selection maps rather than using seed regions as hard masks (ablation) | Geometry-aware network with estimated depth/normals · ScanNet | pixel ℓ1 error: 0.184 -> 0.167 | Helped |
| Use ground-truth depth instead of estimated depth (ablation) | Geometry-aware network with learned selection maps · ScanNet | pixel ℓ1 error: 0.167 -> 0.159 | Helped |
| Fine-tune the entire network end-to-end after stage-wise training (ablation) | Geometry-aware network · KITTI and ScanNet | pixel ℓ1 error | No clear effect |
Closest papers by meaning
- Novel View Synthesis with Pixel-Space Diffusion Models (CVPR 2025 · similarity 0.90)
- Know Your Neighbors: Improving Single-View Reconstruction via Spatial Vision-Language Reasoning (CVPR 2024 · similarity 0.88)
- G-NeRF: Geometry-enhanced Novel View Synthesis from Single-View Images (CVPR 2024 · similarity 0.88)
- Less Decoder is More Encoder: Geometric Representation Learning from Novel View Synthesis (NeurIPS 2026 · similarity 0.88)
- LVSM: A Large View Synthesis Model with Minimal 3D Inductive Bias (ICLR 2025 · similarity 0.87)
- Efficient-3Dim: Learning a Generalizable Single-image Novel-view Synthesizer in One Day (ICLR 2024 · 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.