Research map / 3D vision / Depth and stereo
Long-Range Grouping Transformer for Multi-View 3D Reconstruction
Liying Yang, Zhenwei Zhu, Xuxin Lin, Jian Nong, Yanyan Liang
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 |
|---|---|---|---|
| use long-range grouping attention for inter-view communication instead of per-view-only self-attention (ablation) | LRGT encoder · ShapeNet 20 views | IoU: 0.7682 -> 0.7742 (LGA without IFS) | Improved |
| add inter-view feature signatures (IFS) on top of LGA (ablation) | LRGT encoder · ShapeNet 20 views | IoU: 0.7742 -> 0.7786 | Improved |
| add skip connection combining deconvolution and transformer features in HR basic unit (ablation) | LRGT decoder · ShapeNet 3 views | IoU: 0.7573 -> 0.7590 | Improved |
| use progressive upsampling decoder instead of prior decoders (EVolT, LegoFormer, 3D-RETR) (ablation) | LRGT · ShapeNet 3 views | IoU: 0.7511-0.7576 -> 0.7590 | Improved |
| use full-range attention across all views' tokens (ablation) | Multi-view transformer encoder · ShapeNet with 16-20 views | IoU | Hurt |
| use token-range grouping attention instead of no inter-view attention (ablation) | Multi-view transformer encoder · ShapeNet | IoU | Hurt |
| use long-range grouping rather than short-range or random-range grouping (ablation) | LRGT encoder · ShapeNet, varied views | IoU | Improved |
| use LGA and IFS on single-view Pix3D (ablation) | LRGT · Pix3D single-view | IoU: 0.304 -> 0.299 | Hurt |
| train with 8 views (LRGT+) instead of 3 views | LRGT · ShapeNet 1 view vs 20 views | IoU: 1 view: 0.6962 -> 0.5847; 20 views: 0.7786 -> 0.7922 | Mixed |
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.
- UMIFormer — UMIFormer: Mining the Correlations between Similar Tokens for Multi-View 3D Reconstruction (ICCV 2023) · better on 5 of 7 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
- Multi-View 3D Reconstruction With Transformers (ICCV 2021 · similarity 0.88)
- Block-Sparse Global Attention for Efficient Multi-View Geometry Transformers (CVPR 2026 · similarity 0.86)
- Understanding Multi-View Transformers (NeurIPS 2026 · similarity 0.86)
- TurboVGGT: Fast Visual Geometry Reconstruction with Adaptive Alternating Attention (NeurIPS 2026 · similarity 0.84)
- A Global Depth-Range-Free Multi-View Stereo Transformer Network with Pose Embedding (NeurIPS 2024 · similarity 0.84)
- GTA: A Geometry-Aware Attention Mechanism for Multi-View Transformers (ICLR 2024 · similarity 0.84)
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.
