Research map / 3D vision / 3D generation
MVGBench: a Comprehensive Benchmark for Multi-view Generation Models
Xianghui Xie, Jan Eric Lessen, Gerard Pons-Moll
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 |
|---|---|---|---|
| fine-tune SV3D on same objects (ablation) | SV3D · GSO | cPSNR: 26.751 -> 27.708 | Improved |
| add RCN camera embedding (ablation) | SV3D fine-tuned · GSO | cPSNR: 27.708 -> 28.257 | Improved |
| add CaP camera embedding (ablation) | SV3D fine-tuned · GSO | cPSNR: 27.708 -> 27.980 | Improved |
| combine RCN and CaP embeddings (ablation) | SV3D fine-tuned · GSO | cPSNR: 28.257 -> 27.809 | Hurt |
| replace CLIP with ConvNextV2 encoder (ablation) | SV3D+CaP · GSO | cPSNR: 27.980 -> 28.934 | Improved |
| replace CLIP with DINOv2 encoder (ablation) | SV3D+CaP · GSO | cPSNR: 27.980 -> 28.507 | Improved |
| add SyncDreamer 3D convolution synchronization (ablation) | SV3D+CaP+ConvNextV2 · GSO | cPSNR: 28.934 -> 28.881 | No clear effect |
| increase training objects from 10k to 150k (ablation) | SV3D+CaP+ConvNextV2 · CO3D (real) | cPSNR: 24.261 -> 25.994 | Improved |
| increase training objects from 50k to 150k (ablation) | SV3D+CaP+ConvNextV2 · GSO (synthetic) | cPSNR: 28.053 -> 28.934 | Mixed |
| pass GT multi-views through SV3D autoencoder (ablation) | GT multi-view images · GSO30 | cPSNR: 31.94 -> 30.29 | Hurt |
| use per-object oFID instead of dataset-level FID | image quality metric · user study | agreement with human rankings: 0.50 -> 0.77 | Improved |
| use SV3D-based video model instead of Zero123 | multi-view generator · GSO (best setup) | cPSNR: 17.37 -> 26.75 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SyncDreamer — SyncDreamer: Generating Multiview-consistent Images from a Single-view Image (ICLR 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MVGenMaster: Scaling Multi-View Generation from Any Image via 3D Priors Enhanced Diffusion Model (CVPR 2025 · similarity 0.90)
- MVD-Fusion: Single-view 3D via Depth-consistent Multi-view Generation (CVPR 2024 · similarity 0.89)
- MET3R: Measuring Multi-View Consistency in Generated Images (CVPR 2025 · similarity 0.89)
- ViewDiff: 3D-Consistent Image Generation with Text-to-Image Models (CVPR 2024 · similarity 0.88)
- Ouroboros3D: Image-to-3D Generation via 3D-aware Recursive Diffusion (CVPR 2025 · similarity 0.88)
- SimVS: Simulating World Inconsistencies for Robust View Synthesis (CVPR 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.
