Research map / 3D vision / 3D generation
SAIL-VOS 3D: A Synthetic Dataset and Baselines for Object Detection and 3D Mesh Reconstruction From Video Data
Yuan-Ting Hu, Jiahong Wang, Raymond A. Yeh, Alexander G. Schwing
Check 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 |
|---|---|---|---|
| Use temporal mesh prediction from the previous frame (ablation) | Video2Mesh · SAIL-VOS 3D, All | AP mesh: 9.66 -> 10.21 | Improved |
| Initialize using a class-specific mean shape (ablation) | Video2Mesh · SAIL-VOS 3D, All | AP mesh: 8.96 -> 9.23 | Improved |
| Estimate a transformation between the mean shape and current object (ablation) | Video2Mesh · SAIL-VOS 3D, All | AP mesh: 9.23 -> 9.66 | Improved |
| Pretrain the mesh head on SAIL-VOS 3D | Video2Mesh transferred to Pix3D · Pix3D, 10% training data | AP mesh: 9.0 -> 16.4 | Improved |
| Pretrain the mesh head on SAIL-VOS 3D | Video2Mesh transferred to Pix3D · Pix3D, 20% training data | AP mesh: 16.8 -> 22.5 | Improved |
| Pretrain the mesh head on SAIL-VOS 3D | Video2Mesh transferred to Pix3D · Pix3D, 50% training data | AP mesh: 36.6 -> 38.0 | Improved |
| Pretrain the mesh head on SAIL-VOS 3D | Video2Mesh transferred to Pix3D · Pix3D, 100% training data | AP mesh: 43.7 -> 42.8 | Hurt |
| Use Video2Mesh rather than MeshR-CNN | 3D mesh reconstruction · SAIL-VOS 3D, All | AP mesh: 9.68 -> 10.21 | Improved |
| Use Video2Mesh rather than MeshR-CNN | 3D mesh reconstruction · SAIL-VOS 3D, Heavily Occluded | AP mesh: 3.34 -> 3.62 | Improved |
| Use Video2Mesh rather than MeshR-CNN | 3D mesh reconstruction · SAIL-VOS 3D, Long Clips | AP mesh: 9.03 -> 9.26 | Improved |
| Use Video2Mesh rather than the strongest reported baseline | 3D mesh reconstruction · SAIL-VOS 3D, Medium Objects | AP mesh: 10.73 -> 9.78 | Hurt |
| Use Video2Mesh rather than the strongest reported baseline | 3D mesh reconstruction · SAIL-VOS 3D, Short Clips | AP mesh: 13.39 -> 12.85 | Hurt |
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.
- Pixel2Mesh† — Pixel2Mesh: Generating 3D Mesh Models from Single RGB Images (ECCV 2018) · better on 6 of 8 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
- SAIL-VOS: Semantic Amodal Instance Level Video Object Segmentation - A Synthetic Dataset and Baselines (CVPR 2019 · similarity 0.88)
- Online Adaptation for Consistent Mesh Reconstruction in the Wild (NeurIPS 2020 · similarity 0.87)
- EA3D: Online Open-World 3D Object Extraction from Streaming Videos (NeurIPS 2025 · similarity 0.86)
- Stereo4D: Learning How Things Move in 3D from Internet Stereo Videos (CVPR 2025 · similarity 0.86)
- SPAR3D: Stable Point-Aware Reconstruction of 3D Objects from Single Images (CVPR 2025 · similarity 0.85)
- Fin3R: Fine-tuning Feed-forward 3D Reconstruction Models via Monocular Knowledge Distillation (NeurIPS 2025 · similarity 0.85)
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.
