Research map / Architectures and efficiency / Attention and transformers
IncVGGT: Incremental VGGT for Memory-Bounded Long-Range 3D Reconstruction
Keyu Fang, Changchun Zhou, Yuzhe Fu, Hai Helen Li, Yiran Chen
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 |
|---|---|---|---|
| apply composite-view registration and top-k cache pruning | StreamVGGT · Inference time, 100 frames | time (s): 20.09 -> 8.20 | Improved |
| apply composite-view registration and top-k cache pruning | StreamVGGT · Inference time, 300 frames | time (s): 136.72 -> 25.50 | Improved |
| apply composite-view registration and top-k cache pruning | StreamVGGT · Attention operators, 300 frames | operators (k): 4299.78k -> 126.33k | Improved |
| apply composite-view registration and top-k cache pruning | StreamVGGT · Energy, 300 frames | energy (J): 23516.1 -> 811.0 | Improved |
| apply composite-view registration and top-k cache pruning | StreamVGGT · Reserved memory at 1k frames | memory (GB) | Improved |
| apply composite-view registration and top-k cache pruning | StreamVGGT · Video depth: Sintel | Abs Rel: 0.328 -> 0.341 | Hurt |
| apply composite-view registration and top-k cache pruning | StreamVGGT · Video depth: Bonn | Abs Rel: 0.059 -> 0.064 | Hurt |
| apply composite-view registration and top-k cache pruning | StreamVGGT · Video depth: KITTI | Abs Rel: 0.173 -> 0.176 | Hurt |
| apply composite-view registration and top-k cache pruning | StreamVGGT · Multi-view reconstruction: 7Scenes | median Acc: 0.0241 -> 0.0266 | Hurt |
| apply composite-view registration and top-k cache pruning | StreamVGGT · Multi-view reconstruction: NRGBD | median Acc: 0.0520 -> 0.0516 | No clear effect |
Closest papers by meaning
- Rethinking the State Update Gate for Long-Sequence Recurrent 3D Reconstruction (NeurIPS 2026 · similarity 0.90)
- LiteVGGT: Boosting Vanilla VGGT via Geometry-aware Cached Token Merging (CVPR 2026 · similarity 0.89)
- TurboVGGT: Fast Visual Geometry Reconstruction with Adaptive Alternating Attention (NeurIPS 2026 · similarity 0.88)
- FrameVGGT: Coherence-Preserving Memory for Bounded Streaming Geometry (NeurIPS 2026 · similarity 0.88)
- VGGT: Visual Geometry Grounded Transformer (CVPR 2025 · similarity 0.87)
- BAT3R: Robust Online 3D Reconstruction with Bayesian Adaptive State Updates (NeurIPS 2026 · 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.
