Research map / 3D vision / Face and shape
Novel Views of Objects from a Single Image
Konstantinos Rematas, Chuong H. Nguyen, Tobias Ritschel, Mario Fritz, Tinne Tuytelaars
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 synthesized novel views to 26 side-view training images | DPM, side views · PASCAL VOC 2007 car test | average precision (%): 16.2 -> 30.2 (N=3) | Improved |
| add synthesized novel views to full VOC training set | DPM, full VOC · PASCAL VOC 2007 car test | average precision (%): 51.7 -> 50.2 (N=3) | No clear effect |
| add synthesized novel views to full VOC training set | DPM, full VOC · PASCAL VOC 2007 car test (rare viewpoints) | average precision (%): 53.1 vs 52.5 -> 57.3 (N=4) | Improved |
| add synthesized novel views to side views | DPM, side views · PASCAL VOC 2007 car test (rare viewpoints) | average precision (%): 10.3 -> 32.9 (N=5) | Improved |
| add synthesized novel views to side views | DPM, side views · UCLA cars test | average precision (%): 41.5 -> 83.0 (N=3) | Improved |
| add synthesized novel views to full VOC training set | DPM, full VOC · UCLA cars test | average precision (%): 75.4 -> 84.0 (N=3) | Improved |
| add synthesized novel views to full VOC training set | DPM, full VOC · UCLA cars test (rare viewpoints) | average precision (%): 69.8 -> 81.4 (N=3) | Improved |
| use rendered 3D model instead of real car in training | DPM · PASCAL VOC 2007 car (pilot) | average precision (%): 15.4 -> 11.5 (N=1) | Hurt |
| use re-synthesized car instead of real car in training | DPM · PASCAL VOC 2007 car (pilot) | average precision (%): 15.4 -> 15.0 (N=1) | No clear effect |
| add fine alignment stage with on-the-fly rendering to coarse alignment (ablation) | 2D-to-3D alignment · Pascal3D subset (80 images) | alignment performance ratio to reference: 5% average improvement, roughly 5 times faster | Improved |
| use final coarse sampling scheme (mask, view, envmap, roll variation) over exemplar (ablation) | 2D-to-3D alignment · Pascal3D subset (80 images) | alignment performance ratio to reference: 2-3% improvement, approx 40 fold speedup | Improved |
| add pruning by user mask bounding box (ablation) | coarse alignment · per-class samples | samples tested: 633 out of 6400 samples tested | Improved |
Closest papers by meaning
- Novel Object Viewpoint Estimation Through Reconstruction Alignment (CVPR 2020 · similarity 0.89)
- ViVid-1-to-3: Novel View Synthesis with Video Diffusion Models (CVPR 2024 · similarity 0.88)
- The More You See in 2D the More You Perceive in 3D (CVPR 2024 · similarity 0.87)
- Novel View Synthesis with Pixel-Space Diffusion Models (CVPR 2025 · similarity 0.87)
- Efficient-3Dim: Learning a Generalizable Single-image Novel-view Synthesizer in One Day (ICLR 2024 · similarity 0.87)
- Free3D: Consistent Novel View Synthesis without 3D Representation (CVPR 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.
